The big news story this cycle is that OpenAI secured two large contracts with two separate DRAM companies at the same time, carefully timing the deals so that neither DRAM company knew about the deal with other. Had either company known about the true demand from OpenAI they would have charged a lot more, but each only saw about half.
In other words, there was no collusion between the DRAM manufacturers. They were both caught off guard and left a lot of money on the table.
The current price increase is the result of the huge demand spike. Production takes years to ramp up, but demand has spiked rapidly. Supply and demand.
I find it very telling that both Samsung and SK Hynix already stated that they don't plan to expand capacity - officially to prevent overcapacity in the future. It would also be plausable that both doubt OpenAI will follow through with the contract.
Expanding manufacturing capacity takes many years. Memory has historically been a cyclical business with boom and bust periods. It’s reasonable for manufacturers to be cautious about deciding to expand.
If the demand holds I’m sure they’ll expand. Until then, I think they see it as a short term supply spike.
They don't need to expand capacity to fulfill this contract.
They would want to expand capacity if they believed this increase in demand is long lasting - the implication is therefore that they don't believe it, or not enough to risk major capital expenditures.
You saw the same with GPU makers not wanting to expand capacity during the Cryptocurrency boom. They don't want to be left holding the bag when the bubble pops.
Part of that equation, FWIW, is that certain countries would flood the market with supply to make any new projects suddenly unprofitable.
Which sucks extra bad because if you shut the project down but start it back up you can't just flip a switch. Gotta put together a whole new team and possibly retrain them.
I sincerely hope that OpenAI goes down in flames with those DRAM contracts are going to the highest bidder then so probably Google or whatever AI competitor.
Honestly having problems remembering the other AI companies without googling it. I recall MS, Facebook, Amazon, Google, and Anthropic.
When OpenAI fails, all the AI projects everywhere else will also be killed. Such is the nature of bubbles. With any luck the farm land where the datacenters were built will be sold back to the farmers for half off and they get a free barn out of the deal.
It's more likely that overcapacity is put to work in a plan B, like cheap cloud virtual desktops. Why spend effort on spying and tracking users when their whole desktop computer is in your data center?
When the AI bubble pops, it's going to take the good parts of AI out with it.
Capitalism has never been about the survival of the fittest. That's just weird Nietzschean-Libertarian fantasy where someone ends up blaming the lack of truly free markets for their inability to get a date.
Any large building in a rural area will be used as a barn if it has no other useful purpose. It's kind of hilarious when I pass by the old AT&T long lines facility being used as a hay barn.
I refuse to read the AI slop that passes for journalism about the contracts OpenAI bought, but if they take physical delivery and open actual datacenters built with the RAM they'll be parted out at the minimum, if not absorbed by another AI provider or Big Tech in general.
> prevent overcapacity" is just a fancy way of saying "we prefer to gouge consumers at little risk to us."
No it’s not. Memory business has been cyclical for years. Over expansion is a real risk because new manufacturing capacity is very expensive and takes a long time to come online.
If they could make new manufacturing come online quickly they would do it and capture the additional profit of more sales.
If you present an operating profit of €25 billion USD, yes, in a healthy true market competition would force you to either A) eat into your profit margin by reducing prices or B) invest in R&D and capacit-
Actually, let me eat my words, you are right. As I typed this I saw some news from an hour ago[0] about SK Hynix planning to invest about $500 billion into 4 more fabs. I imagine [hope] Samsung will follow, and together with Chinese memory fabs ramping up both in capacity and technology, prices will return to earth in 2027, maybe 2028.
Guess I am just a little too bitter because GPU prices finally seemed to normalize after half a decade of craziness. Topped with corporations in the West usually forgoing investment and using profits like these to do massive stock buybacks and dividends, souring my expectations.
Additional profit? They're making a lot more money right now than if they had more supply.
The risk of overexpansion is real but I really doubt they want to expand much in the next couple years. They don't have to worry about being undercut by small competitors so they can enjoy the moment.
No they are making higher margins, but not getting as much profit as they could have.
Look at the standard Econ 101 supply-demand curve.
If they could make and sell twice as many chips, it would not cut there margins anywhere near half. They would be making much more.
When demand spikes up and down there will be pain. Because booms are not predictable, in timing, size or duration. And accelerating supply expansion is very expensive, slow, and risky.
Many boom prompted RAM supply expansions have ended in years of unprofitable over capacity.
Price spikes like we are seeing reflect tremendous pent-up/increased demand.
Any price increase reduces purchases by many customers. This tends to keep prices stable. With only small changes in price relative to regular changes in demand.
Yet prices have gone way up.
Which means that many people and businesses are cancelling, delaying, or scaling back their RAM purchases. And yet new demand is incredibly high.
To get prices down, supply would have to grow tremendously. Enough to soak up even more purchases from the very motivated, and to cover all the purchasers that have currently pulled back.
There's room for making more, but I don't think doubling makes sense from a profit point of view.
Especially because the demand curve that's skyrocketing right now is the RAM that isn't in long-term contracts. Doubling all production would much more than double the RAM available for normal purchases.
> To get prices down, supply would have to grow tremendously. Enough to soak up even more purchases from the very motivated, and to cover all the purchasers that have currently pulled back.
Is "down" here back to normal levels?
But normal levels are like a tenth of the profit margin. They'd make significantly less money doing that.
Now thinking of this from the other side, 2 big DRAM producers are taking the risk to dedicate a very big part of their production to AI and if we assume they also have similar deals with other AI companies or big datacenters, what is their risk profile if the AI bubble bursts? Are they viable as companies then ? What is their plan B ?
Their risks are none. They are not increasing capacity, only selling the available one to the highest bidder. Whenever these AI companies run out of money, these producers can simply resume their regular business.
It only depends on whether they get addicted to the high prices.. as long as they can withstand a collapse in the prices then you're right they have minimal risk
Not sure about DRAM companies, but many businesses would still go under if they sold their annual production to a company that then goes bankrupt and won't pay anything for the delivered goods.
Hopefully they get paid more than once a year. Their risk is completely dependent on 1. the net X days until they are paid, and 2. How fast they delay shipment when/if a payment is delayed.
> what is their risk profile if the AI bubble bursts?
Exactly. This is why they’re not scrambling to invest in additional capacity. If these memory manufacturers went all in on new capacity it would take years to build out. If the bubble bursts, or even if it doesn’t burst and just tapers off back to normal demand, they would be in a bad position with excess manufacturing capacity that isn’t paying off.
I think the price increases we are seeing are a direct result of the skepticism about AI scale viability. The big dram houses aren’t increasing capacity, due to the risks you mention.
So demand from other sources has to be suppressed through being priced out in order to meet those supply promises made to OAI in ignorance of their true scale.
This is OAI doing suppliers dirty by making economy distorting moves without transparency, intentionally distorting the market in an effort to hurt competitors.
Yet another example of the “free market” creating destruction for the general public.
As a thought experiment, replace “dram” with “rice” or another essential food stock. Market manipulation such as this is wildly irresponsible, anti-humanity and antithetical to public good. Wars are started over less.
This is an excellent example of the actual alignment of OpenAI as an organization. Yet we are to trust them with leading the way in the alignment of our manque oracles of truth and power?
> This is OAI doing suppliers dirty by making economy distorting moves without transparency, intentionally distorting the market in an effort to hurt competitors. Yet another example of the “free market” creating destruction for the general public.
At the speed OpenAI is growing, it's far more likely they're trying to protect themselves first, not harm competitors. The market only exists because it's free / semi free. Were it controlled by statist bureaucrats - which is the sole alternative back in reality - the situation would be drastically worse. Just ask Soviet Russia. You'd get your meager once every ten year DRAM ration and you'd like it.
The general public isn't the standard of morality or good. Invoking it is meaningless.
I think we can dispense with the strawman Soviet Russia alternative lmfao.
In a reasonably well regulated market, deception at that scale (that utterly destroys competitive buildouts by externalizing the costs that normally would be borne by a customer needing an exceptional order) would be a clear violation of market laws. The fact that deceptive, aggressively anticompetitive behavior such as this , blatantly harmful to other innovation passes as “free market” is a laughable assertion… this is merely the will of the stronger, not any reasonable definition of a free market. A free market implies transparency in pricing and demand, alongside fair competition practices.
Anyone else planning to innovate in the ML space just took a huge hit thanks to OAI, including scientists, pharmaceutical companies, and other things that arguably operate mostly in the realm of clear public good.
Their inherent assumption that might = right is a very powerful indication of their inability to be trusted in the control of a tool / weapon that has more potential to steer the future of humanity than nuclear power/weapons ever did. It’s clear that A: they don’t see AI as any big deal, or B: they don’t care how their actions affect humanity in any nuanced sense of the concept.
> The big dram houses aren’t increasing capacity, due to the risks you mention.
Except they are
> SK hynix to boost DRAM production by a huge 8x in 2026, still won't be enough for RAM shortages
> It's also not just SK hynix that is boosting DRAM production capacity, with both Samsung and Micron rapidly increasing their respective DRAM production numbers.
That's such an impossibly big number for that timeline. The actual news is they're ramping up their newest node, which they were doing anyway, and which was a small percent of their total production.
lol 8x in 2026 hahahahahahaha that is one of the funniest things I’ve ever heard of coming from a semiconductor manufacturer. Maybe 8x as many of something they weren’t selling beforehand, but increasing production on full fabs by 8x? I’d love to be wrong but this makes zero sense to me.
You mean OpenAI will profit selling their RAM stocks if the AI bubble bursts? I doubt it honestly. If the AI bubble bursts, then global demand will collapse altogether crashing the value of HW.
Never gonna happen. Cash has no intrinsic value exccept maybe for use in fire / toiler paper. GPUs while currently inflated in price will always find enough value. Their price might go down 50-75% but never 99%
That's the meaning of intrinsic calor - the device can do what it can do, regardless of market conditions. Today it has the value of fifty teraflops, and tomorrow it still does, unless it breaks. However, intrinsic value cannot be measured in dollars.
And yet we're talking about electronics here, they don't have sentimental value and just because compute capacity is unused there are no guarantees that it will be used, even at a per unit cost approaching €0.
I'm sure that farmers during the Great Depression were also consoling themselves with the "intrisinsic caloric value" of their corn.
As I said, the intrinsic value of a GPU is not measured in €. In fact, the lower the sale price gets, the better a deal it is, not worse - you get the same intrinsic value for less extrinsic cost.
There are also intrinsic costs, mostly power consumption.
Food calories are cheaper to convert into something useful. It's not like GPUs, once bought for peanuts, turn into perpetual motion machines. They need power, cooling, a whole infrastructure built around them.
GPUs would have taken the world by storm already in the roughly 30 years since they've been around.
Even for GenAI it's likely ASICs take over at some point if we really care about performance.
GPUs used to cost 20% of what they cost know and Intel and AMD make perfectly serviceable GPUs for most PCs. NVIDIA top of the line GPUs won't suddenly be plugged in to lowly laptops.
Yes, lots of companies will buy them for cheap, but these AI beasts also have OpEx costs. Not every alternative use is worth the money and there are 0 guarantees that the alternative costs cover the gap. NVIDIA sell 80% of GPUs for AI now.
I think people don't realize just how big this bubble is.
After GPU crypto mining became unprofitable Chinese manufacturers took "mining only" cards, desoldered the GPU and built new graphics cards using the chips.
So at least the lower end stuff (RTX6000) could be repurposed like that.
I find it very hard to believe that two South Korea based Chaebol memory producers would have no wind of this from the other.. more likely is that they both agree that it is not in their or the nation's interest to take on too much of a risk profile... while also balancing the risk of appearing to be colluding.
It's more advantageous long term for them to be oblivious to it. Ultimately gives them what they want which is reduced supply and increased pricing for them.
One difference that strikes me with the .com bubble is that I don't remember the .com companies having sustained multi-billions losses / cash burn. They were not profitable but this is quite different. If (or when) the music stops, won't OpenAI go bust immediately? That's quite a counterparty risk those companies are taking.
Cisco was very much not "stuff like pets.com". Most of the money lost in the dotcom crash wasn't in pets.com, it was the infrastructure companies like Cisco and Sun.
I believe we'd have to tease out what proportion of that cash burn is essential to keep serving compute to customers (which I assume to have profitable unit economics), versus what percentage is optional datacenter buildouts that could be paused in this situation.
Massive cash burn was an absolutely key feature of the dotcom boom/bust. Admittedly, it never really went away - there's always been a free->enshittification profit taking cycle since then. It's just the scale that's terrifyingly different this time.
And when demand eventually crashes, all the new production capacity is left without buyers to sell to, so maybe it does not even make sense to create it.
You have to wonder whether they realy intend to use the ram, or they just spotted an opportunity to corner an essential market and choke/extort in resale.
> Right now it seems like these wafers will just be stockpiled in warehouses – like a kid who hides the toybox because they’re afraid nobody wants to play with them, and thus selfishly feels nobody but them should get the toys!
I'm wondering this as well. Buying 40% of global production just sounds too much. What kind of user counts would they require for that much compute to pay off? Billions of people? What's the chance they could actually get that many users and charge them money? Zero?
Do they have the extra money to play games like that? I thought they were pulling back on ads because of the community reaction even though they wouldn't be introducing it now unless they needed it.
They are not going to repeat the Windows 8 debacle again are they? (Over producing in response to perceived demand only to not get the demand when they are ready to sell) Didn't that bankrupt some memory manufacturers?
Fake demand - fueled by ponzi debt financed by other debt.
Oh, forgot to mention - fake supply too. China was doubly-excluded from the RAM market by using both equipment export bans and tariffs - ensuring the supply is frozen solid.
I'm pretty sure this isn't a coincidence or due to incompetence.
Between inflation generally, and DDR4 being obsolete and unsupported by current desktop or server CPUs, it would be unsurprising for DDR4-3200 DIMMs to be at an all-time high even without the current DRAM price shock. You can never count on old memory types dropping to bargain prices, because the major manufacturers are always eager to migrate the bulk of their production capacity to current-generation memory.
Here the price hike was pretty instant as secondary effect of DDR5 evasion in two waves. I July and now in October.
There is usually no shortage of old working PC components as they also are avalable used and tested from people decommissioning and upgrading systems.
These are not some rare parts in normal market situations.
I made a habit of maxing out motherboards a year or two before upgrading to an new platform. This was always dirt cheap until like 5 years ago.
Yeah the cheapest time to buy old tech is always just when the new stuff has come out. That's when suppliers are trying to shift old stock at cheaper margins.
You can take a look at the 5800X3D and how it was at its cheapest about 2 years ago when AMD was winding down production and Zen 4 had been launched.
To further your point, a used 5800X3D still goes for ~$300 when you can get a brand new 7800X3D for the same if not slightly cheaper (was on sale at MC for $280). I assume the high cost of DDR5 is pushing more people to not upgrade to AM5 and stay on AM4 for as long as possible. And most people are avoiding Intel chips out of principle after the chip degradation debacle - you can see that based on how much lower Intel motherboards are even though they have considerably better feature sets than the AMD equivalents.
Main reason for Intel boards being cheap is because they're practically one and done because of Intel's insistence on changing sockets every other generation.
I wonder if semi-reliable RAM could be made to work for training. After all gradient descent already works in a stochastic environment, so maybe the noise from a few flipped bits doesn't matter too much.
Well, it kind of depends. With XMP (which is overclocking) I've found plenty of kits on Ryzen not passing memtest with the XMP settings. Different CPUs seem to be able to run their memory controller harder without error.
And then there are other factors like more sticks of ram stressing things further. I had to downclock to get memtest stable when running 4 sticks even though each kit ran fine on it's own. But that is expected as well as 4 sticks stresses the memory controller even further.
I confess I don't have any real recent experience with DDR5 though, mostly with DDR4 on Ryzen 1000-5000 series.
Maybe it's time we make a simple web page 100KB again?
Is there some kind of CDN minification, adblocking and compression service?
Maybe even server side rendering of websites?
Then a smartphone would work fine with 1GB of RAM and everyone could be happy.
I've come to the opinion that for the vast majority of apps I've built, it could all be built using HTML + CSS (all built server side). I can sprinkle in little bits of interactivity using something like HTMX. And I'll have a website that is very easy to optimise, has phenomenal backwards compatibility, and gets rid of a whole class of issues associated with SPAs.
I often regret in my career not pushing back more on "requirements" that ended up requiring a more complicated app, whereas the customer would have been happier with a simpler solution.
I guess you're right, but it's more of a curve, though. Once you get to any decent level of complexity, it actually helps to have a framework instead of just going all HTML+CSS. Also it helps having something standard as react (that every web developer should fundamentally understand) than doing your custom stuff if other people will be working on it in the future.
There's a lot to say about the side effects of frameworks but there's a reason why everything converges towards that.
I think it's the other way around, a framework will get you up and running quickly, but then it becomes technical dept, and if your app is complicated you will end up fighting the framework.
If you write something from scratch it will take a while to reach to the abstraction level where you can work fast. But then you have a fully custom abstraction layer that is not a "one size fits all" but custom tailored for your needs.
Good luck with hiring, onboarding, and maintenance of your bespoke solution. Also with your resume when seeking your next gig. For any serious project, ignore community and ecosystem health at your peril. To be clear, we're talking about framework selection, not leftpad vs DIY.
I agree, most websites allow complex flows.
But I suspect, that most loads don't ever touch those.
There is probably an automated way to deliver just a flat page, and maybe even allow for the top 5 interactions without loading all the frameworks and libraries.
Apple could make it happen. For some reason when an iPhone won't load something, people blame the thing instead of the phone. If they made Safari show an error page when a page used more than 256MB of RAM, suddenly the problem would disappear overnight.
No. Safari is already so buggy that it often crashes with the “a problem repeatedly occurred” message. That same message is shown when a webpage requests too much RAM. The problem did not disappear. People rightfully blame iPhone.
I was a bit shocked the other day when Chrome gave me a complaint about memory shortage and said I had something like 200GB in website data on my 256GB laptop(!) All put there by various web pages without my knowledge.
Most people have phones that can handle webpages with 1-5MB JS bundles. Why artificially limit what you can do on the web? Why limit ourselves to 1GB RAM when more resources means tech becomes more useful?
Returning to simple webpages is popular idea on HN but it’s like wanting a car with no backup camera and crank windows. If your goal is to have your car be as simple as possible, then sure, but that’s not the case for most people.
Most people want their cars to be safe and convenient, and their webpages useful and rich, more so than they want to return to some idealized simplicity.
A simple webpage or blog with minimal styling that runs as an ARM binary on a TV remote is cool and fun but it’s not economically useful. It’s the equivalent of a manual scooter. We can build better apps (in the same way that car manufacturers can build less crappy infotainment systems) but optimizing for scarcity isn’t the answer in a world where abundance tends to grow.
(Edit: your downvotes mean nothing to me, I’ve seen what gets upvoted!)
I’d argue that in many of these instances, less is far more.
I want my car to just be really good at being a car, reliably get me from A to B. A Bluetooth connection to the stereo system is nice, but I don’t need a freaking 20” phablet right next to my face when I’m driving.
When I go to a website, I’m usually looking for information, to read something. I don’t often want fancy scroll and animations, I just want clear readable text free of distractions.
More and more these two examples seem to be going away, we’re losing the plot of what the point of these things are.
In a lot of ways, I agree with you. I think the key thing is that the complexity should be appropriate to what needs to get done.
Animations and etc. that distract from the actual content are superfluous. Agreed! I hate it when sites scrolljack.
But lots of HN posters want to impose the same austerity on every website, regardless of whether it’s appropriate. You can’t build Linear in 100KB of JS. Nor would you want to run it on 1 GB RAM. And that’s the case for a lot of economically useful applications.
Keeping things as simple as possible shouldn’t be the goal. It should be keeping it simple enough for the use case at hand.
Your mistake is assuming there is some correlation with usefulness and size.
The JS Gmail UI from 15 years ago was just as functional as the one today.
Websites that are supposed to be simple lists end up bloated and laggy because of really poor JS that makes one request per item iteratively to populate a list.
I do like the old JS Gmail UI. But the current JS Gmail UI doesn’t feel any slower. It is cluttered with more features, but some of them I find useful. (Displaying my calendar and being able to accept invites right in Gmail being a big one.)
As someone who used the HTML gmail interface right up until google pulled the plug: the JS version is much slower to load. Every morning, I get to have about 10 seconds thinking about how it used to be faster.
It absolutely is slower. To an extreme degree even. It takes 10 to 20 seconds to load and is incredibly sluggish to use on some low-end machines I use regularly.
I disagree that we should be optimizing for low end machines and holding back on product improvements for the 80% mass market. Technology improves, it’s one of its best traits. We don’t need to be stuck in the past.
I think those machines are super fun and a snappy Linux experience is very satisfying. I use a lightweight WM myself on Linux and prefer it over the heavy ones. But this segment is 0.1% of technology users and we shouldn’t constrain applications to the limited hardware that this population chooses to use.
I’m glad you want that! But, most people wouldn’t. Also, electric seat adjustments give you way more options than the manual adjustments could. And typically with more precision than the under seat bar with discrete positions.
with you on manual seats? But crank windows? Nah man, power windows and locks are a requirement for me, as is a modern sound system, ac, and cruise control.
You can do a lot with little, it just requires investing more in development which understandably most companies are uninterested in. Besides, plenty of websites are bloated as all hell. Why does a newspaper website, for example, have to be very much more than plain html?
Newspaper websites are a good example of bloat, true. I think if you’re in the business of primarily serving text content and not doing much interactive stuff, you don’t need a heavy site. A lot of them tend to cram their websites with trackers and ads and I guess that’s a business thing.
Tbh, it’s unpopular around HN, but I felt like AMP was a great experience for users. AMP pages were super fast and had no annoying banners - and none of my pet peeve: layout shift.
I just checked my go-to stores in Europe. Holy shit. I bought a 5200-speed 16GB DDR5 kit in October for 55 eur. Now it's selling for 240 eur. The 6000-speed model is now 390 eur.
You actually get more, for the 1500€ you don't have to mess with sticks yourself. They solder it directly on your mainboard. If that isn't great service, I don't know what is.
I've been meaning to get more components for a diy NAS since atleast the last year and just been pushing it lazily. I'm literally kicking myself now when I actually started looking up deals for this black friday.
Same here, I was waiting for DDR5 to come down in prices to upgrade from 32 -> 128gb. Could have done it for 200-300€ a few weeks ago, now I'm just sad.
Depending on how many disks you need, buying old PCs on eBay is the way to go because there are still some Xeon E3 ECC models available cheap ($100) with plenty of RAM (16GB) and four SATA ports for a NAS.
On October 1st OpenAI signed two simultaneous deals with Samsung and SK Hynix for 40% of the worlds DRAM supply... the shock wasn’t that OpenAI made a big deal, no, it was that they made two massive deals this big, at the same time, with Samsung and SK Hynix simultaneously! In fact, according to our sources - both companies had no idea how big each other's deal was, nor how close to simultaneous they were. And this secrecy mattered. It mattered a lot.
Had Samsung known SK Hynix was about to commit a similar chunk of supply — or vice-versa — the pricing and terms would have likely been different. It’s entirely conceivable they wouldn’t have both agreed to supply such a substantial part of global supply if they had known more...but at the end of the day - OpenAI did succeed in keeping the circles tight, locking down the NDAs, and leveraging the fact that these companies assumed the other wasn’t giving up this much wafer volume simultaneously…in order to make a surgical strike on the global RAM supply chain…and it's worked so far...
OpenAI isn’t even bothering to buy finished memory modules! No, their deals are unprecedentedly only for raw wafers — uncut, unfinished, and not even allocated to a specific DRAM standard yet. It’s not even clear if they have decided yet on how or when they will finish them into RAM sticks or HBM! Right now it seems like these wafers will just be stockpiled in warehouses – like a kid who hides the toybox because they’re afraid nobody wants to play with them, and thus selfishly feels nobody but them should get the toys!
The cynic in me thinks this would be a convenient way for these memory producers to manufacture demand, while also making OpenAI look good on paper. It’s not like they haven’t been caught price fixing in the past. Win win for these companies and a loss for everyone else.
On paper this makes OpenAI look like absolute assholes. Like they have realized that all of their potential competitors will be memory constrained and have poured billions of dollars into making sure that happens instead of using that money to improve their own product.
If manufacturing fake demand (warehousing 900,000 memory wafers per month?) doubles and triples customer prices, it would decrease real demand, hurting the manufacturers after the artificial demand ends.
I think I explained that poorly. Basically artificially reducing supply so that these manufactures can get more for less. They've been caught doing it in the past before between each other, so why not use OpenAI as a bridge for that.
Fake demand? Either they’re selling the RAM or they’re not. They don’t make money by pretending to sell into fake demand. They make money by selling chips. A sale is a sale.
> Had Samsung known SK Hynix was about to commit a similar chunk of supply — or vice-versa — the pricing and terms would have likely been different
Wouldn't this be ... collusion?
Implicitly arguing that the memory oligopoly should have been coordinating is ... quite something.
OpenAI may well be doing something anticompetitive by cornering supply to foreclose competitors, but saying "they tricked the suppliers into not colluding!" is certainly a take you can have I guess.
No, it's not collusion to ask for more money from OpenAI if you hear that they are trying to buy 40% of the world's supply. Increased demand leads to higher prices, that's normal.
OpenAI, by doing simultaneous deals, hid the true demand from the suppliers, thus lowering their price and raising everyone else's.
There is nothing suspicious or abnormal about this behavior. It is called competition. Ironically, trying to prevent this kind of behavior prevents competitiob, and is a key factor for causing monopolization
It depends... if OpenAI bought the DRAM in order to use it, then fair play to them.
If they bought the DRAM in order to stop their competitors from using it because they are falling behind, that's anticompetitive in spirit, though I'm not sure if it actually breaks any laws.
Knowingly attempting to buy or sell in quantities likely to move markets, for direct profit, is called manipulation and is most definitely illegal. this is true in physical markets commodity markets and financial markets. Not saying that this is what openAI is doing but it definitely merits an investigation.
All quantities bought or sold on the margin will move the market. Whether it moves the market is not up to the buyer or seller; it is up to other buyers and sellers who react to that transaction and adjust their expectations. This is normal market dynamics, dynamics we should want to happen because markets adjusting to movements of big players performs a social function; you and I need to know how large movenents of resources affects our livelihoods, and this is how that can happen.
There is no reason to pathologize or find suspicious these normal economic facts. Especially when it is not within the power of a big player to choose how other people react to their actions, which is all "moving markets" is. If something is suspicious and illegal about that, then it is equally suspicious that you and I seem to go along with this "market movement" by these big players and pay the new prices. Are we colluding with them? We could do with less conspiracy-minded interpretations of these things.
market economics are like newtonian mechanics. It's all so wonderful and logical and even elegant, until the dimensions expand a few orders of magnitude, and then all the rules break. Having worked on a trading floor for 20 years I know how this works. Swamping a market with huge trades is definitely considered manipulation by essentially all authorities, and indeed is a form of monopoly power, which even economic theorists will agree is undesirable. Jane Street just got a mega fine for exactly this in India, btw.
I can't get over the confident dismissal of science by hand-waving about imperfect modelling. But what I said has nothing to do with that, and is more true to the real world than an idealized perfect competition model. Pathologizing normal trading behavior like this is more the result laymen and authorities misinterpreting bad economic modelling. So I recommend you take some of your own medicine and look at the mirror. Maybe a trade affecting the market isn't so suspicious as you make it out to be, because the perfect competition model you're using to make accusations of monopoly simply doesn't make sense. Again, if there is something wrong with affecting the market, then you or I are just as liable for our consciously self-interested behavior of choosing higher-quality, lower-priced products.
If you know demand will go up because Microsoft announced that each new Xbox will have 2TB of RAM, that is perfectly fine. Or if OpenAI issues a press release that they intend to buy half the worlds RAM.
If you know demand will go up because you learn the volume your customer intends to purchase from your competitor during confidential negotiations, that is not ok.
Every normal market interaction could be described by scary monopoly-sounding words. You heard about you're friend's salary in the same industry, and ask for a higher salary next time you negotiate. You're colluding with your friend to price gouge your employer!
In reality, so-called collusion is normal and unobjectionable. But when price surges happen, often due to factors outside of the seller's immediate control, people look for any reason to find an ethical dimension and find how to place blame, because this is more convenient. Things that were normal become abnormal and suspicious. It is in consumers economic self-interest to act in this way, because it often secures favors to them from various economic policies that they don't normally get when the market is "normal". This is no less a form of collusion than what sellers might do to secure their economic advantage. But a key difference for these anti "gouging" policies is that it gives consumers a special privilege and makes market pricing less able to fulfill its social functions.
Only 50-60% of the recent price appreciation is after October 1st, so this is not likely the biggest direct cause, more likely it's everyone buying the finished stock on the market in a frenzy because they anticipate things like this will impact availability in the future, and it becomes a self-fulfilling prophecy.
Sounds like OpenAI might be trying their hand at TPUs, like what Google has. They are one of Google's biggest advantages in AI right now. It would also give them insurance against Nvidia being everybody's hardware supplier.
What do you mean “might be” trying their hand? It’s been widely reported that they are doing exactly that. It’s even known that they are doing it with Broadcom.
This actually points the opposite direction, to doubling down on commercial GPUs.
NVIDIA recently told their board partners that they will need to source their own RAM and will not be bundling it with chips anymore.
If there is a supply crunch on DRAM, commercial GPU production lines will start having idle downtime. That is literally the worst possible thing that can happen to a company that has invested heavily in tooling and they will negotiate at or below cost production runs to fill the gaps if a customer can bring their own DRAM to the table.
Novice question: If they built something other than classic dram modules with the wafers maybe they could achieve faster bus speeds? How does Apple do it?
Apple uses off the shelf LPDDR modules. They have nothing whatsoever special about them.
Apple gets high bandwidth out of these modules not with high bus speeds, but with a very, very wide bus. This is expensive on the SoC side (requires a large die, which is why others don't necessarily do this), but allows for commodity memory modules.
That's what HBM is actually. The memory dies are directly next to the GPU die, on the same substrate. The main difference between Apple SoC and GPUs is the former use regular LPDDR while GPUs use HBM.
One of the key points of HBM is that dies are stacked up with many, MANY, more signals and channels. That's how NVIDIA has a memory bandwidth an order of magnitude higher than M4: 550GB/s for the M4 Max, 4.6TB/s for H200. And yes, that's bytes per second, not bits per second.
some GPUs use HBM. Most use GDDR. AMD and Nvidia still extract huge bandwidth from GDDR via high bus speeds + wide buses (like the 1.79 TB/s on the 5090)
Indeed! I implied "AI GPUs" since that’s where HBM is commonly used (despite AMD pioneering it on some consumer cards). And yeah, thousand-bit wide busses get close in performance.
shorter traces than soldered DIMM allow higher MT/s, this is fixed by CUDIMM/CAMM2, the other part of this is # of memory channels on the board, not sure why, but most consumer DDR5 boards have been 2 memory channels, you need to go to threadripper to get 4 or 8, It's unclear to me if this will still be an issue with future platforms.
You're not paying enough attention to the performance and cost impact of connectors and sockets. CAMM2/LPCAMM and CUDIMM have yet to be demonstrated operating at speeds that speeds that match the fastest soldered LPDDR, let alone GDDR; there's still a clear advantage for soldering memory.
CPU sockets with more than two memory channels are also far more expensive; the higher pin count usually increases the number of layers the motherboard needs, and the larger size of the socket requires more metal for stiffening (and EPYC CPUs still have issues with imperfect mounting leading to some IO lanes not working).
Using BGA soldering for both the processor and the memory sidesteps a bunch of engineering challenges.
> Using BGA soldering for both the processor and the memory sidesteps a bunch of engineering challenges.
by trading them with longevity challenges.
Even though 2 decades have past after the "usual suspect" lead-free solder, gpu or vram chips needing a reball is still a common occurrence from a cursory look at YouTube channels of professional electronics repairmen.
That's not blogspam. There's a lot more in the mlid post than the linked news item above. Whether it's commentary or speculation I urge everyone to read and decide for themselves.
When Gaben's 11 organize a heist to steal the RAM wafers from the high-security vault in Sam Altman's basement, the price of the steam machine will drop below $1000 USD.
As someone currently dealing with pricing changes for DDR5 volume contracts (admittedly, only ~30k DIMMs, but still), those contracts are a lot less rigidly priced than we (engineering) realized at the time - “agreed price” is a lot more... flexible... until the pallet is delivered. Especially because our contract is with our manufacturer who has their own contract with a DIMM manufacturer who has their own contract with the DRAM manufacturer. DRAM is substantially more like a spot market than the average consumer would expect (at least at my volumes). The same is true for my HDDs (~100k unit/year contract) and my CPUs (10k unit/year contract). At least for HDDs our contract quantities are getting honoured (and we actually still have Seagate and WD fighting for our business), which I’ve heard hasn’t been the case for smaller orders.
Yep. It's a bit like some gas projects when prices spiked due to Russia's attack on Ukraine.
Companies had supply agreements in place but they will find essentially any excuse to "delay" delivery. You might eventually get your product but fat lot of good it does for you while you're mired in court and not getting anything anyway.
The particularly amusing example is a project continuously selling spot LNG while saying it's not yet operational, stiffing the companies with which they'd signed long term contracts with on a technicality.
For a while, yes. But I'm sure they will want to sell it for longer than a year or so.
Even Samsung is running into this issue now: their own internal foundry is refusing to give them a long term contract now so the S26 series will become more expensive.
If this happens to Samsung, what leverage will a player like Valve have?
> Even Samsung is running into this issue now [...]
These large corpos are so greedy to the point they harm themselves. I remember something similar with Amazon, where the Amazon shop had to redo the whole architecture from some microservice setup back to a monolithic approach because they used AWS and paid these weirdly structured prices like everyone else. Which made running a monolithic architecture under such constraints inherently cheaper. Not sure what the resulting compounded business cost for this "endeavor" was, but more often that not such things are never accounted for, so they don't show up as an issue.
That's their costs, not our costs. Regular PCs will cost more, which means Valve can price their consoles higher and collect margin, because there's no other competition/substitutes.
Maybe, but from a platform and brand development perspective, there may be smarter strategies.
Their interest isn't necessarily in squeezing out the most margin on each unit. If unexpected market conditions let them more or less hit their original margin targets but get far more units installed in homes, that could be much better for them in the long term.
There's a lot we can't know as outsiders, at the moment.
> that could be much better for them in the long term.
i agree. Valve's money machine is with the steam platform, rather than any hardware sales - breaking even on hardware is sufficient imho.
Valve's existential threat is from microsoft closing off windows somehow (or extracting rent...like a store!). Therefore, pushing steam machines which can be run without windows, is both business expansion as well as an insurance policy.
I'd wonder how much Valve has agreed to buy other parts and tooling/manfacturing services. Then they could have scenarios where they have a pile of AMD chips and the rest of the parts, but limited or no VRAM/RAM/NAND to complete the systems so they can be sold, or they've spent a pile of money designing and working with a manufacturer to get to the stage where they can start making batches, but can't get the volume to bring in a real benefit.
Being pragmatic, I wonder if Valve benefit from the health of the PC market, whether they'll also put efforts into helping people get more life out of their existing machines, or individuals/businesses refurbishing to sell on the used market with more confidence. There's already an angle for reducing e-waste with the win10 end of life if it can be tied into installing linux.
> i agree. Valve's money machine is with the steam platform, rather than any hardware sales - breaking even on hardware is sufficient imho.
Wasn't that true with the previous incarnation of the steam machine, the valve index, steam controller, etc.? IIRC their VR gear was some of the most expensive consumer level gear on the market.
Those controllers are first gen stuff - for the enthusiasts. They're expensive but not unduly so, not to mention that VR is still a niche, and is unlikely to expand the steam store. The first steam machine (steam box iirc) had both production issues and also not had much enthusiast success. Likely due to low hardware specs and not targeting the console market.
The steam deck, and now the (newer) steam machine is looking to push into the mass market of consoles, and i reckon it would be more successful. Not to mention the more mature steamOS and compatibility layer now (compared to before).
I have high hopes for it. I want PC gaming to break free of microsoft windows.
I upgraded my pc about this time last year (new CPU, RAM, and motherboard).
I was originally going to just get 64GB of DDR5-6000, with the option of adding another 64GB later, thinking the price might drop even further. At the last minute, I decided to get the whole 128GB instead. Glad I did.
I don't understand this. I'm looking at prices on some Asian store fronts and it's nowhere even remotely near these. I'm looking at DDR5-6000 2x16 for about $130, with no apparent limits.
Even with tariffs, transport, and other fees, you could get this to the US for way less than $400. I doubt the market could be this inefficient - in other words I don't think I just found a get rich quick scheme. So, what gives?
The current prices are a response to future stock, not current. It's 100% retailers price gouging with their current stock that they got for cheap because they know there will be limited stock in the near future. Asian retailers may be more honest and keep their margins the same, but will catch up in a month or two.
It's just surge pricing to manage demand, not "price gouging". Being able to buy RAM at high prices is a lot better than being unable to source it at all because everyone else is panic-hoarding.
Price gouging has a specific meaning. It’s when costs are raised rapidly only because people are desperate and have no other choice. (e.g. water after a disaster)
Retailers raising prices in reaction to an incoming event that will take supply away from everyone is not that.
We don’t really know whether that’s true, since it’s hard to prove a negative (i.e., suppliers aren’t colluding). But given their history of price fixing it may be worth looking into.
Is it better? As with all price gouging, better for those who can afford it, sure, but not for those who can't. The proper way to combat scalping is to implement fair allocation methods (for a start purchase quantity limits) and punish people for scalping.
Look at how most places handled war-time gasoline shortages. Rationing coupons, purchase limits, demand leveling (like the odd-even system), price or profit controls, strict prosecution of scalpers and price gougers. And it's not like only the communists did this - even the US had most if not all of these things. And it worked far better than the shit that happened during the pandemic shortages. Governments used to know how to govern.
If you want to be "fair" for a necessity such as gasoline, you can have tradable rationing coupons. That way you are rewarded if you buy and use less gasoline, but the excess windfall due to the shortage is still transfered to you and away from the supplier. But even this assumes that gasoline is in fixed supply and there is no way of increasing its total production by paying more, which is not a very good assumption.
In a time of shortage, throwing more money at the problem usually won't increase supply. A shortage necessarily means that if you make more of something, you're guaranteed to sell it basically instantly, so there's already an incentive to increase production.
And it's not like the higher prices mean more money goes to the producers so they can invest in more production capacity. The price increase is spread out between every middleman in the chain untils there's almost nothing left. This could work only if the producers themselves are the ones raising prices, but then everyone else would still add their own cut, leading to even crazier price hikes, and also it's unlikely that extra profit would go to much more than lining the owners' pockets.
Additionally, demand spikes usually don't last, so any new production capacity you build will be a liability later, after the market settles down.
>A shortage necessarily means that if you make more of something, you're guaranteed to sell it basically instantly, so there's already an incentive to increase production.
This is patently false. Every oil reserve around the world has a cost per barrel of extraction. At $60/barrel many of them are shut down.
If you fix the price at $60 and demand goes up, you’ll end up with shortages and producers won’t be able to fill the gap.
It is very rare to have a market of physical goods where the cost of production is fixed and supply is effectively limited. For every other market, the price needs to go up to entice investing in making more supply.
The high prices ("price gouging") perform a social function because some cannot afford it; they prioritize what matters to them more, and ram is left available to those who absolutely require it. Trying to get around that by forcing prices below the market price simple encourages scalping behaviors. If prices are below the market price, but at the market price you would prefer the cash in your pocket over the ram stick, then you have every reason to sell it higher than you bought it because people are willing to buy it. It is those willing to buy it that are the main culprit in establishing the true price.
In reality, it is almost never a true binary of "afford" or "cannot afford" like critics of surge pricing make it out to be; people evaluate the price according to their circumstance and make a trade off. It is because of these decisions, the state of demand, that surge pricing is possible, not because of the machinations of evil price scalpers. That is why manufacturers couldn't lower prices even if they wanted to; gpu msrp being a great example of gpu vendors being caught between consumer ignorance about economics and the facts of reality that gpus are scarce enough to warrant higher prices.
I'm definitely not saying it's a binary can/can't afford situation. The point is that people don't "afford" things equally. I might some need RAM so badly that I'm prepared to take a huge risk and spend half of my paycheck on it. But that doesn't matter because someone else has 3x my paycheck, savings, and investment portfolio and a good credit score. How much money someone can spend on somethong is no indicator of how much they need/want it.
Something like "GPUs are actually scarce" doesn't even make sense to say, since scarcity is more a function of demand than supply. The supply of GPUs wasn't exhausted because people suddenly needed more GPUs or because Taiwan couldn't produce as many of them as they used to, it was because a few rich bastards were buying into a bubble so they could make as much money a possible before it all comes crashing down. They didn't "need" those GPUs much more than even the scalpers. They were just a vehicle to make short-term profits at the expense of everyone else.
And yes, of course those willing to buy things are the ones enabling the peice gouging. But that's not a useful observation. You either need something, so you'll buy it even if it doesn't make financial sense, or it makes financial sense to buy it, so you will. Notice how scalpers also fall into that second category, along with the rich bastards draining the supply.
Scarcity is a function of demand. That doesn't change the importance of what I said! In fact, it proves my point. The reason that "some rich bastard" is willing to pay such prices for gpus or ram or anything else is because they expect it to generate revenue for them equal to or greater than the amount they paid for it. How do they generate revenue? By selling a product! Who do they sell the product to? People like you and I who also compete with them to buy ram and gpus. In other words, the importance of what I said about prices depending on demand is that it depends on your demand. Rich bastards don't care much about ram per se; they could get it anyway. They care about it because of what (they expect) you're willing to pay for its products. Prices are what they are because people, on net, would rather spend the equivalent they could spend on ram by buying products of ram.
Putting this in view to the idea that people don't "afford" things equally: by your assumption, this implies people can indeed "afford" other non-ram things better when it comes to the more important alternatives they could buy with the ram-equivalent funds. Not only do ram-equivalent funds compete with alternate uses, but ram as a factor of production competes with other factors of production. And all factors of production compete, by way of the so-called rich bastards, for your and my dollars. In other words, if ram is more expensive, it is to support alternate uses of ram whose products are valued more highly by consumers than the direct use of ram. And, most importantly if one were to try to get around this higher resulting price for ram, it would cause higher prices for the products of those alternate uses of ram. People would be less able to get the thing they value more highly than ram because ram competes with all our needs, and less ram can be used for its indirect use.
All of that is to say that efforts to combat so-called price gouging bounds those who can less afford ram to be in a worse spot than otherwise. They can't afford ram as before, that much is true. But they prefer the alternatives to ram. If they would be better off by having ram, they would purchase that. Waving a magic wand to redistribute ram to them will give them ram, but now they lose what they valued more highly than ram.
Your logic that OpenAI can by proxy afford to buy out the entire world's supply of RAM because consumers value OpenAI products more than other RAM-dependent things assumes that OpenAI's money comes from selling goods or services to those consumers. It doesn't. The overwhelming majority of people pay zero dollars per year on AI services, while almost everyone spends at least a few hundred a year on gadgets that need RAM and other services that run on servers that need it as well.
The money OpenAI is using to starve the rest of us of RAM is coming from pumped up valuation through circular investments, investor FOMO, cheap debt and often straight up gambling. Rich bastards know that they can pump money into the bubble to grow it and hopefully cash out before it bursts. Nowhere in that process did any regular person value AI datacenters over other uses of RAM.
What you're pointing to is uncertainty; it is a judgement call whether or not consumers actually value these things. But if they don't open AI will lose money, assuming the market is not interfered with by bailouts, which would indeed hamper the social function of pricing. In a similar manner, this social function is hampered by interventions which manipulated credit and deby. My assumption only works in an economy where debt isn't artificially created ad hoc by the federal reserve and other banks. But that is a problem with how federal policies condition debt and finance in the modern economy, not with surge pricing. I would like to see more criticisms of the fed which essentially bankrolls bubbles like this rather than surging pricing, which performs its social function either way.
>since scarcity is more a function of demand than supply
This is incorrect. Anything that is bound by something like TSMC production is only made scarce when nvidia realizes they could sell out the entire run at $1000/card or whatever.
They were supply limited for years during the crypto boom. The way you know it was a supply problem is that you couldn’t even buy new cards because they were so frequently sold out.
Nvidia cards became really valuable overnight in the same way as any other asset. You trying to scream at “rich bastard” buyers will not change the fact that there is a shortage of cards so the price is going to go up across all sellers until the supply and demand curves intersect.
This is basic econ in action and history is fraught with attempts to try to fix supply shortages by capping the price.
> The high prices ("price gouging") perform a social function because some cannot afford it; they prioritize what matters to them more, and ram is left available to those who absolutely require it.
This is not true at all. It isn't left available to those who absolutely need it but to those who can pay for it. Those are two very different things.
You're attempting to construct a notion of need separate from the empirical experience of people's concrete choices, but that isn't possible if you want to be scientific. If a person spends $400 on a phone instead of 32gb of ram, they need the phone more than the ram. If a company spends $400 on ram instead of on some other production good, they need the ram more than the other production good - they expect it to generate more revenue.
These two exchanges are not disconnected either: phone prices are affected by revenues of companies which use ram for production. And those revenues are determined by purchases of phones. The person demonstrates through their choice that the use of ram for an indirect purpose of making phones (however indirectly that might effect it) is more valuable to them than the direct use in their computer. The person is not excluded from "having" the ram in the most general sense: they have it indirectly because they benefit from its use in production whose products they value more than the direct use of ram. The person, along with all other consumers, participates in organizing production in the manner that best benefits them, according to their needs, which may not necessarily involve them directly owning the thing.
During the pandemic there wasn’t gouging? I just remember complete shortages of inventory like toilet paper, basic microchips, etc.
It would have been better if people did raise prices during the pandemic for those things to prevent hoarding so I could actually wipe my ass at 2 cents a wipe instead of 1. But alas, the “price gouging” cry babies would have come out and lambasted them for “being greedy”.
Nope, governments were famously bad at this. Coffee rations and gas rations were a disaster.
(Edit: Replying here because of dumb rate limits)
>You guys are being unreasonable, we have plenty of toilet paper for everyone. Each person gets two rolls per week unless you can prove you need more until you calm the fuck down
And this is why it’s dumb. There actually was a supply shortage. You should read about it.
Toilet paper manufacturers made industrial scale toilet paper for offices, schools, public buildings, rest areas, etc.
1/3 of the entire toilet paper market for giant single ply rolls sold in bulk disappeared overnight. And that same demand flowed back into home multi-ply toilet paper that couldn’t be scaled up quickly because it came from a different mill.
Rations would have been completely stupid in reaction to a legitimate 50% increase in legitimate demand.
Read again. My whole point was that governments used to know how to handle a shortage, but they don't anymore, as evidenced by the pandemic. They let the market figure it out and you got a mix of both price gouging and scalpers.
And no, doubling prices wouldn't have done anything. Hoarders would just hoard more because not only was supply low, but prices were increasing, so they better buy it now rather than later.
If governments actually governed, this wouldn't have been a problem. "You guys are being unreasonable, we have plenty of toilet paper for everyone. Each person gets two rolls per week unless you can prove you need more until you calm the fuck down. We're also putting the toilet paper factory into overdrive to compensate for this stupidity."
It's speculative price gouging. Calling it "surge pricing" doesn't stop erosion of consumer trust in the market. Watch now as more people more readily jump to price fixing conclusions. Not helped by the inevitable further increase in speculation through feedback loops and the resultant volatility.
First come first served is a better principle than "surge pricing". A lottery is a better principle than "surge pricing". In the case that someone over purchased, they're free to dispose via secondary market if the value to them is lower than the out of stock price. I.e. decentralised pricing (and profits). Secondary market sales are just more efficient, they occur at negotiated prices that reflect true individual valuation, not the retailer's speculation.
I'd rather reward diligence and personal responsibility - if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price. Rather than passive reliance on wealth to solve problems. First come first served values effort and foresight. Scarcity is managed through time and effort rather than money.
> First come first served is a better principle than "surge pricing".
This is called a price ceiling, and it's a bad idea with a track record of failure and significant harm.
I'd rather pay extra and get what I need with 100% chance than get what I need cheaply with 5% chance and otherwise be forced to go without or buy from scalpers for the same price I would have paid anyways. This is the purpose of prices. So the people who really need it can buy it, and those who are borderline about the purchase decide to opt out.
If you're concerned with wealth inequality or one large buyer cornering the market, there are better ways to address those problem than prices ceilings.
The act of eliminating surge pricing is not a price ceiling. That's a different thing. That requires more than simply swapping surge pricing with first come first served. You've created a strawman.
> I'd rather pay extra and get what I need with 100% chance
False dichotomy. Neither approach increases supply. Of course according to economists who can hand wave away bullwhip effects with simple "this model assumes X" statements that go unquestioned in the conversations which cite the findings of the given model but i digress. According to economists, both approaches do increase supply, the theory goes that the price gouging retailer invests in more factory capacity. Or the factory owner buoyed by vibrant secondary market activity views increased production investment as a safe bet. Maybe there's some truth in the latter...
> If you're concerned with wealth inequality
I'm concerned with lazy financial engineering over hard work. Why should the scrappy but innovative startup be excluded from resources over the sclerotic incumbent with a deeper wallet?
> The act of eliminating surge pricing is not a price ceiling. That's a different thing. That requires more than simply swapping surge pricing with first come first served. You've created a strawman.
What is it then? If you allow the price to increase, there is no need to enforce a different rationing mechanism. The only reason to think of implementing the alternate rationing mechanism is if the good isn't being rationed. If you insist on the previous, lower price, then that is a price ceiling in so many words, with the consequence of needing your rationing mechanism. If you allow a higher price, then your alternate mechanism is unnecessary.
The bullwhip effect is the whole reason why retailers may want to hold greater product stock in the first place; to absorb transient demand fluctuations and not have to pass them on in full to the supplier.
And a "scrappy but innovative" startup has an obvious interest in being able to source the DRAM or other goods they require, even at higher prices.
> The act of eliminating surge pricing is not a price ceiling.
If you force companies to price products lower than what they want, then you have a price ceiling by definition.
> Neither approach increases supply.
I didn't mention or allude to supply at all, although it's true that price ceilings also decrease supply (less so if there is monopoly or collusion, but they still do).
I was talking about resource allocation. The person who needs a new system because their previous computer broke will be willing to pay the extra money, but the person who already has a DDR4 system with a 5950x that runs their games well enough will be content to hold off on their AM5 upgrade to DDR5 because the marginal improvement isn't worth the extra $400.
If you have a price ceiling like what you proposed, the person with the DDR4 system may buy the DDR5 that they don't really need 1 day earlier than the person who actually needs the DDR5, creating a misallocation of resources.
(that's an example of the more abstract principle at play).
Theoretical arguments aren't even needed here. The empirical history of price ceilings is there for you to google.
If you didn't know that what you were proposing is a price ceiling, and you thought that I was talking about supply instead of resource allocation, then I mean this with no offence intended but you should study elementary economics before forming confident opinions on the subject. Society is in a vulnerable point with cost of living pressure and we don't need more energy behind these harmful populist ideas.
Price ceiling definition: a government-imposed legal maximum price
My original comment: First come first served is a better <snipped for brevity>
This is not a legal maximum price, this is a legal maximum in the derivative of price.
> I didn't mention or allude to supply at all
>> get what I need with 100% chance than get what I need cheaply with 5% chance
How do you square these two statements? One claims 100% supply certainty when no such thing exists in this context. Without making certain assumptions (unstated, but ludicrous, assumptions are rife in economics discourse), you can't state much of value about which buyer will get the goods in the surge pricing model, you especially cannot say that the buyer with the larger wallet will always win. Think for a second what assumptions you've made to this point in the conversation - you're still down the rabbit hole of price ceilings in the comment chain thus far.
>> The empirical history of price ceilings is there
Not disputed but as per your call out of definition above, not relevant.
To make the point further - the name for a limit on rate of change of price is not a price ceiling, anymore than the 0-60 time of a car is its top speed limit.
>> you thought that I was talking about supply instead of resource allocation
My challenge to you is to name the assumptions you've identified in your reasoning around resource allocation. I'm confident i can point out the deficiencies in your model because that is the nature of models.
>> you should study elementary economics
That's a great idea, a really good follow on from that is to identify logical fallacies you discover i. that process, especially those so accepted that it's not a stretch to say they are underpinning the discipline. A good example of that would be the conjectural origin theory of money but i digress.
> My original comment: First come first served is a better
Retailers are first come first serve. The first customer to buy product gets the next unit of inventory.
If you’re trying to argue that retailers need to be forced to set a retail price for each unit of in-stock inventory and then be forced to sell each unit at that price later no matter what the market rate does in the mean time, that’s an awful idea.
I’m sure you don’t mind when retailers decide to give discounts or put things on sale, do you? If the market price drops, they reduce selling price. Same thing when market rate goes up.
Forcing retailers to price things how you want doesn’t change supply and demand. It would only force retailers to add extra margin into their prices to account for the added risk of the government forcing their hand in sale prices.
> It's speculative price gouging ... if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price
The word for "monitoring market trends, anticipating needs and acting quickly" is, well, speculation. Why should a retailer not be allowed to speculate and hold more product stock when they anticipate a future crunch? In fact, the whole reason prices have become so volatile right now is that this supply crunch was not properly anticipated.
The reason why retailers are not "price fixing" is that price fixing involves setting an artificial ceiling on total production; retailers are not in a position to do this, and there is no evidence at all that DRAM makers are doing this either.
> First come first served is a better principle than "surge pricing"
First come first serve must means enterprising individuals would buy as much as they could afford and then immediately relist it on Facebook Marketplace for the actual market rate.
Thinking that retailers could keep RAM prices low and also keep it in stock for you is irrational.
If your goal isn’t to make cheap RAM available to you, then what even is your goal?
Are you just angry that retailers are making a little extra temporary profit on their in-stock inventory? Is the goal to punish them for behaving rationally and understanding basic economics?
This perspective is pretty silly. If you don't know the price you set it up for auction. The Ebay auction prices in German are in the $400 range for 2x16GB. The retailers aren't particularly expensive. Maybe $10 to $20 more.
>I'd rather reward diligence and personal responsibility - if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price. Rather than passive reliance on wealth to solve problems. First come first served values effort and foresight. Scarcity is managed through time and effort rather than money.
Are you really going to discount the effort made by the manufacturers to produce the product? You know, giving them money is the best way for them to produce more products. A lucrative DRAM market will also make it easier for new competitors to enter the market.
Also, if everyone does what you suggest, then all that will happen is that the price will rise extremely rapidly before the actual supply crunch even happened. That sounds like what is currently happening.
> It's 100% retailers price gouging with their current stock that they got for cheap
The retail price of a product is a function of the market rate at the intersection of supply and demand. The price paid for inventory on the shelf doesn’t matter.
It works both ways. If retailers bought a lot of RAM at high prices and then the market suddenly dropped, they could have to sell it at a loss.
Some people get irrationally angry at this, but you do it too. If you bought a house for $500K and the market went up such that it was worth $700K, you wouldn’t think it was “price gouging” to list it at market rate. You’re just trading an asset for cash at the current price. The price you bought it at is irrelevant to the price you’re going to sell it for.
technically the house analogy isn't that good because there are lots of people who won't sell below what they bought for, even if there are no buyers at this price anymore. the house eventually sells if the seller is forced to sell, but price in such cases isn't necessarily 'correct' either. more liquid markets make better examples.
> technically the house analogy isn't that good because there are lots of people who won't sell below what they bought for
Just ignore the downside part of the house analogy because it doesn’t apply to this current DRAM situation where prices went up.
The key point is that if you buy something and its value goes up, you’re not going to offer it for sale at the same price you paid for it. You’re going to sell it at market rate.
Everyone does this, including retailers. It elicits angry cries of “gouging” from people who want to believe retailers are the cause of the high prices and be angry at them, but the market is big. It’s supply and demand.
The market is inefficient. Also, if you are in a small country, the demand is not there. So the seller will sell with regular prices. The stock will also be very low to export it back to where it came from.
Not only that - I think OpenAI and all those other AI blobs must pay compensation damage to The People for inflating the prices here and skyrocketing costs.
Been planning to build my daughter a gaming PC for Xmas. Feel lucky I found a pre-built with 32GB RAM in a warehouse store on black Friday that hadn't been marked up yet. First pre-built desktop I've bought in decades and it was actually cheaper than I could build myself due to the crazy RAM prices.
If things go the way they are, I wonder how long until buying a used Playstation 4 (8GB DDR5 RAM) at around 100USD a piece to put Linux on it and use as a server in cluster will become viable thing to do? Bring back console clusters!
Guess that’s the flip side of selling sometimes three to four years old hardware with zero price cut (I know they’ve been doing better in the M-series era)—the price also stays there when it’s skyrocketing elsewhere.
> What specifically happened in June to set this off?
Tariffs implemented by this administration:
"Inflation has begun to show the first signs of tariff
pass-through," said Ellen Zentner, chief economic
strategist at Morgan Stanley Wealth Management. "While
services inflation continues to moderate, the acceleration
in tariff-exposed goods in June is likely the first of
greater price pressures to come. The Fed will want to hold
steady as it awaits more data."[0]
The tariffs are bad, but you're simply wrong here - US tariffs do not affect global prices, it's OpenAI buying up production in advance and cornering the market.
> Why did European prices have the same increase then?
Where do Europeans get their DRAM from?
If it is the same handful of companies the US gets their DRAM from, then why would Europeans pay any less? Because the EU is not engaging in the same asinine trade war?
Sounds good in theory, but in practice those same few companies can set prices for markets outside the US to be at/near US prices. It doesn't take much effort for manufacturers to set their prices at or near those of their competitors and rely on an implicit mutually assured destruction[0] understanding.
> Sounds good in theory, but in practice those same few companies can set prices for markets outside the US to be at/near US prices. It doesn't take much effort for manufacturers to set their prices at or near those of their competitors and rely on an implicit mutually assured destruction[0] understanding.
If a company in A sells a widget W to both the EU and the USA, such that a consumer in the EU and the USA pay the same prices even though the USA has a tariff and Europe doesn't, then the company will make a lot more profit per unit selling all their W in the EU and none in the USA.
I'm not at all sure what's happening at any given moment with the USA's tariffs on anything, given the chaos over there. But let's say W is the set of all things relevant to AI data centres. What this means is that all the data centres are now much cheaper to build in Europe rather than in the USA. Data centres can be put just about anywhere, given they're used over the internet anyway. This means that the companies selling W would have all the demand they want for W in the EU, so they could sell all of their supply of W in the EU, so they could get a higher profit margin on all of it.
I'm not sure how much DC investment money is going to which parts of the world, but I am sure that if all the suppliers stopped shipping to the USA because they could sell as much as they could make everywhere else in the world for more profit (and the same purchaser price after tariffs), I would have heard about it.
> If it is the same handful of companies the US gets their DRAM from, then why would Europeans pay any less?
... because tariffs are paid for by the buyer?
Importing memory from Korea to the US means the importer had to pay a tariff. Importing memory from Korea to Europe means the importer does not have to pay a tariff. The company selling the memory gets exactly the same amount of money in either case.
> I'm sure if they didn't keep the prices somewhat similar, you would have a bunch of people in Europe selling RAM to Americans.
I was just about to edit my response to the GP to say the same thing. Let's explore this hypothetical situation a bit further.
Suppose there was a DRAM manufacturer named "Acme DRAM" which decided to have a separate pricing schedule for the EU reflecting the lack of insane US tariffs.
Some enterprising entrepreneur in the EU would establish a company in the country having the least US tariffs and resell Acme DRAM to US companies. Surely this would make money hand-over-fist.
Problem is, the US DoJ does not look kindly on this kind of enterprise:
DOJ also has demonstrated a growing willingness to pursue
criminal charges against companies and individuals involved
in customs fraud schemes such as the purposeful
misclassification of goods, falsifying country-of-origin
declarations, and intentionally shipping goods through
low-tariff countries. Importers of goods into the U.S.
should expect criminal enforcement to accelerate in the
coming months and years.[0]
This would then put Acme DRAM in the crosshairs of an already vindictive and erratic US administration, likely to not only hammer the entrepreneur (see above) but to also include tariff ramifications for Acme DRAM as well.
All of this risk in the pursuit of lower profit margins by definition.
> Some enterprising entrepreneur in the EU would establish a company in the country having the least US tariffs and resell Acme DRAM to US companies. Surely this would make money hand-over-fist.
Re-badging is a thing that some companies do actually do, despite what the DoJ says.
> Importers of goods into the U.S. should expect criminal enforcement to accelerate in the coming months and years.
I am absolutely not a lawyer, but wouldn't "importers" be the USA residents, not the EU businesses doing the exporting?
> Afraid not. This has not happened across the board to all components.
Afraid so. Industry impact from tariffs does not require them to be applied "across the board to all components." See here[0] for more information.
With erratic massive tariff proclamations, counter-tariffs are to be expected. All it takes for companies to inflate prices is to either:
A) be provably impacted by tariffs
B) be opportunistic by being "tariff adjacent"
The net result is directly or indirectly, the tariffs implemented and/or threatened to be so are a significant contributor to electronic component costs.
The problem is that training these models and using these models has required exponentially increasing amounts of memory.
ChatGPT has existed for years and in those years it's userbase and model size has increased tremendously. Not to mention the fact that a lot of competitors have sprung up in the wake of GPT. Including the likes of cloud based open model hosting services.
They're not the only player trying to buy more RAM. Every hyperscaler is also growing.
AWS, MS, Oracle, and Google demand is growing ~100% in 2026. Nvidia demand is growing ~50% in 2026. Supply isn't forecast to improve at all until H2 2026 and it's probably ~all spoken for through at least EOY 2026.
That's from June 2025. It's not relevant to current situation — prices have been rising since October 2025 due to agreements that became public in the same month.
12-24 months is my best guess. I bought some at inflated prices today at a retailer that failed to update to the latest market price by 15%. I'm not waiting that long only to be faced with a dilemma about waiting even longer for DDR6 to become reasonably priced.
Somehow another sign that our world is collapsing.
We have more and more wars, less free trade and international spirit, freedom and privacy is going down the sink, "everything is going down", and now ram is a good example of things that use to be abordable and with price going down, becoming expensive luxury good.
One could say that RAM price is just a special temporary event not related to our down trend but a decade ago major players would not have left a profitable consumer market without new actors coming to fill in.
But now we prefer restricting offer instead of increasing production...
It certainly seems clear that a purely upward trajectory of peace and living standards in the West was not supposed to last. History has its ebbs and flows, I'm sure we have a long way downhill to go, but I suspect there are peaceful times on the other side.
Net energy surplus had to end sometime, we have just decided to speed run it for some reason. Looking forward to sledding down the other side of the carbon pulse!
What could be the potential impact on smartphone and tablet prices in the coming months or years? I am assuming that laptop prices will start increasing next year.
"Downloading RAM" is more feasible than people think.
The seek time of a consumer-grade hard disk is said to be on the order of 10 ms. That's roughly the latency of a very high quality FTTH connection. Meaning that if you run a HDD rather than an SSD, a swap file in the cloud could potentially be faster than a local one (especially when you consider multiple reads/writes that could be done in parallel).
It's not exactly downloading more RAM but decently close to call it that for the joke.
It’s not clear to me how long the contracts last. The Tom’s Hardware article linked below mentioned “in 2025”, so, do we know if this monopolistic DRAM grab is nearly over or only just beginning?
I mean I wouldn't say it's misleading, it just says what the graph contains. It's important that it shows 18 months so you can see how flat it was before
They should get back into the Optane business first. Take the existing Optane memory to a modern PCIe 5 bus, and it will be competitive with lots of uses for DRAM. PCIe 5 gives you basically DDR4-like transfer bandwidth, and the Optane media has better latency and far better DWPD endurance than existing NAND SSD's. It's a total no-brainer.
Man, we can't even buy bare DRAM chips at work. Nobody will sell us anything. The whole US electronics industry has suddenly been put in a chokehold for.... perpetuating the AI bubble?
Man it'd be great if these AI assholes could stop absolutely trashing every single part of society, the economy, and the physical planet at large.
I hate this. PC gaming is my hobby, the only one that’s lasted my whole life. It’s always been there. It’s how I met my wife. It’s how I relax after a long day. It’s how I’ve participated in so many stories that stick with me and given me so many memories.
All of it is being murdered by the AI bros. Before them it was the crypto bros. It’s one thing after the other and I hate it so much.
Just live with the PC you have for two more years. It's probably not a big deal? A moderately capable machine from 5 years ago is still marginally capable.
There is more than a lifetime of incredibly great PC games that run on your existing hardware, and if this is your life's hobby, then paying an extra $100 or so every year or few is a drop in the bucket of your gaming expenses
I can afford it. Other people cannot, and the hobby is driven by a market existing for games. If newer people don’t enter the hobby as others die out, it fades away.
I mean its pretty rare to buy more RAM after completing your PC build + that single PC is going to last you 5+ years. Also mobos usually only have 4 slots in total so its not like its even going to take a lot. I'm rocking 2x48gb sticks and that's plenty for gaming.
The prices are wild tho.
I bought that ram in March 2024 for $384.81. Now it's priced at $1,172.99. LOL
Increased demand for computer components for purposes other than gaming constitutes "AI bros murdering your lifelong hobby"?
PC gaming is not "murdered", it's doing better than ever.
In 2015 there were 3,000 games released to Steam, last year there were 18,000. In 2015 Steam's peak concurrent user count was 8.6 million. This year it's 41 million.
The inflation-adjusted price per gigabyte of RAM has dropped from $3/GB to $2/GB over the last 10 years, even including the recent price hikes.
So spare me the hysterics, your hobby is fine.
And you know what? The increased demand for compute always spurs innovation, so you'll probably get a better computer in the end as a result. You're welcome.
> In 2015 there were 3,000 games released to Steam, last year there were 18,000. In 2015 Steam's peak concurrent user count was 8.6 million. This year it's 41 million.
This is like saying "Spotify's subscriber count grew by 800% over the last 10 years. Music is doing better than ever!"
I'm finding that getting an LGA socket to give me two reliable memory channels is far more valuable than the sticks that go into them. I've got at least two motherboards in use at my home right now with only one working channel of memory. There are sticks just sitting around.
Nothing stops you building an EPYC based workstation. For the last socket you could even get mATX boards from Asrock Rack which were kind of ridiculous - I was very tempted to build my PC around one.
A lot of the goods won't be essily repurposible for consumer or small business work loads.
Imagine if auto manufacturers all refitted their factories and supply chains to produce military vehicles for a war effort. New family cars would run dry, and when the war ended, some folks would figure out make clever use of some surplus military vehicles for street travel and commerce, but most of the surplus would just be shifted to other military markets and family car production would take some time to resume.
AliExpress has been selling 2010 or whatever Intel xeons in dual socket desktop board kits for gaming. they are fairly affordable and hold up to almost modern games so people that can't afford new gen systems.
These products are fascinating. When I looked at them it seemed like they pulled old chipsets from server boards or maybe they found stock somewhere and then combined them with south bridges from things like Z77 boards. All to make use of obsolete server CPUs that were cheap and available. Amazing and wonderful.
Meanwhile you've got OpenAI buying up all the DRAM and maybe just piling it in a warehouse so no one else can have it or they figure out what to do with it. Microsoft recently said they don't even have a power source to plug the ton of GPUs they just bought into so they're also just sitting around collecting dust. What is even happening?
Your point was to post nonsense without caring about whether it was true or false in order to troll someone into responding. Please don't do that again.
I don't think you should comment on this site again if you aren't willing to comment in good faith. I pointed out that your original statement was completely false, and rather than thanking me for the correction, you responded defensively and then attacked me. This shows that you don't care whether what you posted was true or not. That's corrosive to the presumption of good faith we employ here. If you aren't willing to behave properly, you shouldn't be here.
At these prices, it might pay off to start an open-ish, non-profit, DRAM company to produce chips good enough for these purposes. They can sell a certain amount above cost for consumers and small businesses. Full price with guaranteed volume for some, enterprise suppliers.
The profits might be used to port the DRAM to multiple foundries to gradually increase supply. Alternatively, they can shift to produce other components, like VRAM. Make the low-to-midrange accelerators with larger VRAM more available at reasonable prices.
Just speculating. I don't know silicon economics well.
Got real lucky. I was trying to negotiate a guy down on RAM. I wanted a server without as much RAM. In the end, I had to take the 768 GB. It only cost $1k more than 192 GB at the time. Unfortunately I can’t really sell off the excess because I need to substitute down the RAM modules to fill all 12. Ah well.
If we measure the value of the dollar against megabytes of RAM, well, in 01987, the difference between a StarBoard with "2 megs space" and a StarBoard with "2 megs installed" was US$484 (https://archive.org/details/amazing-computing-magazine-1987-...), so a "meg" of RAM cost US$242. Today https://pcpartpicker.com/trends/price/memory/ says "DDR4-3200 2x8GB" costs US$108; if that means 16GiB, as in 16384 "megs", a "meg" of RAM costs US$0.0066 today.
So the dollar's value has increased by roughly a factor of 36700 over those 38 years, averaging 32% per year.
That would be an average yearly inflation of -24%.
Too bad you can't live in DRAM or eat it when you're hungry.
> If we measure the value of the dollar against megabytes of RAM
We should be measuring it by the amount of RAM in a typical household PC in 1987 and today.
Even though a "meg" of RAM costs less than 1 cent today, I can't do anything useful with it.
Even if we are generous and buy a whole $1 of RAM today, it only gets us 150 MB of RAM, which, while infinitely more useful than 1MB, is still completely useless for running a modern OS/Browser.
Math says that's stupid. You can still do everything with a "meg" of RAM that you could do in 01987. In fact, the "meg" of RAM you get now is enormously faster, and you can interface it to other things that you couldn't get at the time, such as SSDs, so you can do strictly more with it.
Economics says you're spending newly abundant resources freely in order to conserve those that are still scarce. Economics also predicts that people will adapt to RAM prices doubling by using RAM more frugally, spending more of other resources to compensate.
They seem to fail to capture a whole lot of things. Supposedly $1 in 2000 is worth $1.88 in 2025. So 88% inflation over the 25 years. Meanwhile the median home price has increased by 150%. Family insurance by 350%. Median college tuition by 225%. Childcare costs have risen by 200%. But sure. We can buy super cheap 65" tvs now. Hurray for us! Literal kings who lived hundreds of years ago couldn't possibly imagine a world with cheap large screen tvs. So the poorest among us should rejoice at the wonders they are able to enjoy while they skip meals and ration their insulin.
A dozen eggs is up over 350%, but a 6-pack of Budweiser is only up 60ish% since 2000. So you know, it all balances out. Maybe drink a few extra cans of Bud with your next meal.
This explores the ideas behind your post: important things, like education and healthcare, have disproportionately risen in price while not-so-imortant thing have gotten less expensive.
The inflation basket only represent a hypothetical average person, who doesn't even exist.
It's more useful to construct multiple separate inflation measures that represent different types of people. Like a "typical renter" inflation figure vs a "typical homeowner" inflation figure. It wouldn't be hard to do and would shine a light on inequality and help explain the rise of populism in certain segments of society.
An even better measure would somehow appropriately normalize the figure by the average disposable income in each of the segments to come up with a figure that measures the felt impact better.
The figure would be negative for wealthy people (who actually benefit from inflation because of asset price inflation) and positive for poor people (whose disposable income mostly goes to rent).
You can go one step further than that and calculate a fairness measure using something like the Gini coefficient (*) and analyze how much it has changed over time.
What a ridiculous statement. We have a massive number of people in the USA relying on financial support for purchasing food. Not every single American is fat, especially those who literally can’t afford to eat.
There are lots of Americans who go hungry, but mostly outside the USA. Starvation deaths inside the USA are almost entirely secondary to medical conditions that interfere with eating.
Generally there is a hidden starvation (I'm not sure the exact term), when one consumes just enough calories to survive a bit longer but almost no vitamins and minerals to live normal life.
Similar to polluted air/environment such deaths will hardly be backtracked to the true roots.
The most recent were deaths in conjunction with COVID. Was it due to COVID directly, indirectly or it was so bad that COVID could not really make it worse...
Funny how many dot-com esq things are popping up.
https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
The big news story this cycle is that OpenAI secured two large contracts with two separate DRAM companies at the same time, carefully timing the deals so that neither DRAM company knew about the deal with other. Had either company known about the true demand from OpenAI they would have charged a lot more, but each only saw about half.
In other words, there was no collusion between the DRAM manufacturers. They were both caught off guard and left a lot of money on the table.
The current price increase is the result of the huge demand spike. Production takes years to ramp up, but demand has spiked rapidly. Supply and demand.
To those who haven't heard how colossal the size of OpenAIs contracts are.
900,000 wafers monthly. Tom's hardware estimates that is equal to 40% of global dram production capacity.
https://www.tomshardware.com/pc-components/dram/openais-star...
I find it very telling that both Samsung and SK Hynix already stated that they don't plan to expand capacity - officially to prevent overcapacity in the future. It would also be plausable that both doubt OpenAI will follow through with the contract.
Expanding manufacturing capacity takes many years. Memory has historically been a cyclical business with boom and bust periods. It’s reasonable for manufacturers to be cautious about deciding to expand.
If the demand holds I’m sure they’ll expand. Until then, I think they see it as a short term supply spike.
They don't need to expand capacity to fulfill this contract.
They would want to expand capacity if they believed this increase in demand is long lasting - the implication is therefore that they don't believe it, or not enough to risk major capital expenditures.
You saw the same with GPU makers not wanting to expand capacity during the Cryptocurrency boom. They don't want to be left holding the bag when the bubble pops.
Oil has been like that as well. High oil prices don't trigger nearly as much drilling as they used to.
https://www.cbsnews.com/news/oil-production-prices-us-compan...
Part of that equation, FWIW, is that certain countries would flood the market with supply to make any new projects suddenly unprofitable.
Which sucks extra bad because if you shut the project down but start it back up you can't just flip a switch. Gotta put together a whole new team and possibly retrain them.
I sincerely hope that OpenAI goes down in flames with those DRAM contracts are going to the highest bidder then so probably Google or whatever AI competitor.
Honestly having problems remembering the other AI companies without googling it. I recall MS, Facebook, Amazon, Google, and Anthropic.
When OpenAI fails, all the AI projects everywhere else will also be killed. Such is the nature of bubbles. With any luck the farm land where the datacenters were built will be sold back to the farmers for half off and they get a free barn out of the deal.
It's more likely that overcapacity is put to work in a plan B, like cheap cloud virtual desktops. Why spend effort on spying and tracking users when their whole desktop computer is in your data center?
When's the last time you bought something at Sears?
Connected up to the grid and water supplies, I daresay there’ll not end up as barns.
An air conditioned barn with space for 100,000 cattle!
The AI projects that make sense will live. Capitalism is survival of the fittest.
That farm land is dead and gone, best we can do is urban/rural decay.
When the AI bubble pops, it's going to take the good parts of AI out with it.
Capitalism has never been about the survival of the fittest. That's just weird Nietzschean-Libertarian fantasy where someone ends up blaming the lack of truly free markets for their inability to get a date.
Any large building in a rural area will be used as a barn if it has no other useful purpose. It's kind of hilarious when I pass by the old AT&T long lines facility being used as a hay barn.
They wont be going to highest bidders if they are under contract to produce at the claimed level, unless what you are referring to is the residual.
I refuse to read the AI slop that passes for journalism about the contracts OpenAI bought, but if they take physical delivery and open actual datacenters built with the RAM they'll be parted out at the minimum, if not absorbed by another AI provider or Big Tech in general.
"prevent overcapacity" is just a fancy way of saying "we prefer to gouge consumers at little risk to us."
Hopefully the Chinese manufacturers ram(p) up rapidly and spike Hynix and Samsung with heavily undercut prices.
> prevent overcapacity" is just a fancy way of saying "we prefer to gouge consumers at little risk to us."
No it’s not. Memory business has been cyclical for years. Over expansion is a real risk because new manufacturing capacity is very expensive and takes a long time to come online.
If they could make new manufacturing come online quickly they would do it and capture the additional profit of more sales.
If you present an operating profit of €25 billion USD, yes, in a healthy true market competition would force you to either A) eat into your profit margin by reducing prices or B) invest in R&D and capacit-
Actually, let me eat my words, you are right. As I typed this I saw some news from an hour ago[0] about SK Hynix planning to invest about $500 billion into 4 more fabs. I imagine [hope] Samsung will follow, and together with Chinese memory fabs ramping up both in capacity and technology, prices will return to earth in 2027, maybe 2028.
Guess I am just a little too bitter because GPU prices finally seemed to normalize after half a decade of craziness. Topped with corporations in the West usually forgoing investment and using profits like these to do massive stock buybacks and dividends, souring my expectations.
[0]https://www.pcgamer.com/hardware/memory/hot-on-the-heels-of-...
Additional profit? They're making a lot more money right now than if they had more supply.
The risk of overexpansion is real but I really doubt they want to expand much in the next couple years. They don't have to worry about being undercut by small competitors so they can enjoy the moment.
No they are making higher margins, but not getting as much profit as they could have.
Look at the standard Econ 101 supply-demand curve.
If they could make and sell twice as many chips, it would not cut there margins anywhere near half. They would be making much more.
When demand spikes up and down there will be pain. Because booms are not predictable, in timing, size or duration. And accelerating supply expansion is very expensive, slow, and risky.
Many boom prompted RAM supply expansions have ended in years of unprofitable over capacity.
> If they could make and sell twice as many chips, it would not cut there margins anywhere near half. So they would be making much more.
You really think that? I would expect their margins to drop down to a small percentage if they doubled production. Maybe even less.
Price spikes like we are seeing reflect tremendous pent-up/increased demand.
Any price increase reduces purchases by many customers. This tends to keep prices stable. With only small changes in price relative to regular changes in demand.
Yet prices have gone way up.
Which means that many people and businesses are cancelling, delaying, or scaling back their RAM purchases. And yet new demand is incredibly high.
To get prices down, supply would have to grow tremendously. Enough to soak up even more purchases from the very motivated, and to cover all the purchasers that have currently pulled back.
There's room for making more, but I don't think doubling makes sense from a profit point of view.
Especially because the demand curve that's skyrocketing right now is the RAM that isn't in long-term contracts. Doubling all production would much more than double the RAM available for normal purchases.
> To get prices down, supply would have to grow tremendously. Enough to soak up even more purchases from the very motivated, and to cover all the purchasers that have currently pulled back.
Is "down" here back to normal levels?
But normal levels are like a tenth of the profit margin. They'd make significantly less money doing that.
at these prices, there are certainly potential customers not purchasing when they otherwise would have
You don't maximize profit by maximizing sales.
Now thinking of this from the other side, 2 big DRAM producers are taking the risk to dedicate a very big part of their production to AI and if we assume they also have similar deals with other AI companies or big datacenters, what is their risk profile if the AI bubble bursts? Are they viable as companies then ? What is their plan B ?
Their risks are none. They are not increasing capacity, only selling the available one to the highest bidder. Whenever these AI companies run out of money, these producers can simply resume their regular business.
It only depends on whether they get addicted to the high prices.. as long as they can withstand a collapse in the prices then you're right they have minimal risk
Assuming they haven't massively changed operations to crank up supply, which seems to be the case, they shouldn't be massively hurt with a price drop.
If this price goes on for a longer period though, I assume that won't be the case.
Well an example would be if they for example took out massive debts on the back of an expected Revenue stream
Not sure about DRAM companies, but many businesses would still go under if they sold their annual production to a company that then goes bankrupt and won't pay anything for the delivered goods.
Hopefully they get paid more than once a year. Their risk is completely dependent on 1. the net X days until they are paid, and 2. How fast they delay shipment when/if a payment is delayed.
e.g. dotcom DSL and CLEC providers, https://www.sfgate.com/bayarea/article/dsl-providers-droppin...
Everything is prepaid, just like you buy RAM online.
> what is their risk profile if the AI bubble bursts?
Exactly. This is why they’re not scrambling to invest in additional capacity. If these memory manufacturers went all in on new capacity it would take years to build out. If the bubble bursts, or even if it doesn’t burst and just tapers off back to normal demand, they would be in a bad position with excess manufacturing capacity that isn’t paying off.
I think the price increases we are seeing are a direct result of the skepticism about AI scale viability. The big dram houses aren’t increasing capacity, due to the risks you mention.
So demand from other sources has to be suppressed through being priced out in order to meet those supply promises made to OAI in ignorance of their true scale.
This is OAI doing suppliers dirty by making economy distorting moves without transparency, intentionally distorting the market in an effort to hurt competitors.
Yet another example of the “free market” creating destruction for the general public.
As a thought experiment, replace “dram” with “rice” or another essential food stock. Market manipulation such as this is wildly irresponsible, anti-humanity and antithetical to public good. Wars are started over less.
This is an excellent example of the actual alignment of OpenAI as an organization. Yet we are to trust them with leading the way in the alignment of our manque oracles of truth and power?
> This is OAI doing suppliers dirty by making economy distorting moves without transparency, intentionally distorting the market in an effort to hurt competitors. Yet another example of the “free market” creating destruction for the general public.
At the speed OpenAI is growing, it's far more likely they're trying to protect themselves first, not harm competitors. The market only exists because it's free / semi free. Were it controlled by statist bureaucrats - which is the sole alternative back in reality - the situation would be drastically worse. Just ask Soviet Russia. You'd get your meager once every ten year DRAM ration and you'd like it.
The general public isn't the standard of morality or good. Invoking it is meaningless.
I think we can dispense with the strawman Soviet Russia alternative lmfao.
In a reasonably well regulated market, deception at that scale (that utterly destroys competitive buildouts by externalizing the costs that normally would be borne by a customer needing an exceptional order) would be a clear violation of market laws. The fact that deceptive, aggressively anticompetitive behavior such as this , blatantly harmful to other innovation passes as “free market” is a laughable assertion… this is merely the will of the stronger, not any reasonable definition of a free market. A free market implies transparency in pricing and demand, alongside fair competition practices.
Anyone else planning to innovate in the ML space just took a huge hit thanks to OAI, including scientists, pharmaceutical companies, and other things that arguably operate mostly in the realm of clear public good.
Their inherent assumption that might = right is a very powerful indication of their inability to be trusted in the control of a tool / weapon that has more potential to steer the future of humanity than nuclear power/weapons ever did. It’s clear that A: they don’t see AI as any big deal, or B: they don’t care how their actions affect humanity in any nuanced sense of the concept.
> The big dram houses aren’t increasing capacity, due to the risks you mention.
Except they are
> SK hynix to boost DRAM production by a huge 8x in 2026, still won't be enough for RAM shortages
> It's also not just SK hynix that is boosting DRAM production capacity, with both Samsung and Micron rapidly increasing their respective DRAM production numbers.
https://www.tweaktown.com/news/109011/sk-hynix-to-boost-dram...
Note to self don't trust tweak town.
That's such an impossibly big number for that timeline. The actual news is they're ramping up their newest node, which they were doing anyway, and which was a small percent of their total production.
Is this new capacity or will some kind of other chip type suffer?
lol 8x in 2026 hahahahahahaha that is one of the funniest things I’ve ever heard of coming from a semiconductor manufacturer. Maybe 8x as many of something they weren’t selling beforehand, but increasing production on full fabs by 8x? I’d love to be wrong but this makes zero sense to me.
This also creates massive risk.
Financial trouble at OpenAI - even minor stuff where they slow purchases by 25% - could have a big impact on global prices.
At that point even if the AI bubble bursts you have a solid business as a RAM scalper
You mean OpenAI will profit selling their RAM stocks if the AI bubble bursts? I doubt it honestly. If the AI bubble bursts, then global demand will collapse altogether crashing the value of HW.
Free GPUs for everyone! Bring a truck!
I honestly can't wait to use used gpus as lego bricks similar to what kids in Weimar republic did with cash
My self-training robot army is poised to conquer all. It's merely waiting for parts.
Never gonna happen. Cash has no intrinsic value exccept maybe for use in fire / toiler paper. GPUs while currently inflated in price will always find enough value. Their price might go down 50-75% but never 99%
I've got an S3 VGA adapter to sell you at 25% of list price
They're taking about shorter time scales, the effect on top of the normal obsolescence treadmill.
> GPUs while currently inflated in price will always find enough value.
What is the instrinsic value of one of millions of GPUs, if the world only needs 15-20% of them?
About fifty teraflops.
That's the meaning of intrinsic calor - the device can do what it can do, regardless of market conditions. Today it has the value of fifty teraflops, and tomorrow it still does, unless it breaks. However, intrinsic value cannot be measured in dollars.
"calor" was a typo of "value" obviously.
And yet we're talking about electronics here, they don't have sentimental value and just because compute capacity is unused there are no guarantees that it will be used, even at a per unit cost approaching €0.
I'm sure that farmers during the Great Depression were also consoling themselves with the "intrisinsic caloric value" of their corn.
As I said, the intrinsic value of a GPU is not measured in €. In fact, the lower the sale price gets, the better a deal it is, not worse - you get the same intrinsic value for less extrinsic cost.
There are also intrinsic costs, mostly power consumption.
Demand for GPU power is much more elastic than demand for food calories.
Food calories are cheaper to convert into something useful. It's not like GPUs, once bought for peanuts, turn into perpetual motion machines. They need power, cooling, a whole infrastructure built around them.
GPUs would have taken the world by storm already in the roughly 30 years since they've been around.
Even for GenAI it's likely ASICs take over at some point if we really care about performance.
GPUs have taken the world by storm. There's one in almost every computer, and they make up the bulk of supercomputers!
If you put a 75% discount on these powerful GPUs there will be a long line of non-AI-company purchasers.
GPUs used to cost 20% of what they cost know and Intel and AMD make perfectly serviceable GPUs for most PCs. NVIDIA top of the line GPUs won't suddenly be plugged in to lowly laptops.
Yes, lots of companies will buy them for cheap, but these AI beasts also have OpEx costs. Not every alternative use is worth the money and there are 0 guarantees that the alternative costs cover the gap. NVIDIA sell 80% of GPUs for AI now.
I think people don't realize just how big this bubble is.
The most likely outcome is kids in third world countries extracting the valuable metals from piles of discarded GPUs.
After GPU crypto mining became unprofitable Chinese manufacturers took "mining only" cards, desoldered the GPU and built new graphics cards using the chips. So at least the lower end stuff (RTX6000) could be repurposed like that.
Soon you will bring a wheelbarrow of GPUs to buy a loaf of bread.
If AI bubble bursts -> global market crash. Ehm, with what you will be scalping, with bottle caps fallout money? :D
I don't really buy that AI crashing will cause "the global market" to crash. How exactly do you think that works?
I find it very hard to believe that two South Korea based Chaebol memory producers would have no wind of this from the other.. more likely is that they both agree that it is not in their or the nation's interest to take on too much of a risk profile... while also balancing the risk of appearing to be colluding.
I wonder if they caught wind, but they both thought they were competing for the same contract.
Another way of making the point I'm making is you don't become one of the two leading memory producing capable companies in the world by being naive
It's more advantageous long term for them to be oblivious to it. Ultimately gives them what they want which is reduced supply and increased pricing for them.
One difference that strikes me with the .com bubble is that I don't remember the .com companies having sustained multi-billions losses / cash burn. They were not profitable but this is quite different. If (or when) the music stops, won't OpenAI go bust immediately? That's quite a counterparty risk those companies are taking.
Dotcom companies were basically stuff like pets.com, not things considered strategic with de-facto government backing.
Cisco was very much not "stuff like pets.com". Most of the money lost in the dotcom crash wasn't in pets.com, it was the infrastructure companies like Cisco and Sun.
Funnily enough Cisco’s stock has recently recovered back to its dotcom peak.
Thanks to the irrational exuberance of the AI bubble...
Cisco did not go bust.
And I didn't say it went bust.
I believe we'd have to tease out what proportion of that cash burn is essential to keep serving compute to customers (which I assume to have profitable unit economics), versus what percentage is optional datacenter buildouts that could be paused in this situation.
OpenAI is ten years old, dotcom companies were 2-3 years old.
Some dotcom-boom companies that survived also had sustained multi billion dollar losses afair - Amazon and Uber for example.
Uber was founded half a decade after the dot com bubble.
"half" confused me but okay 2001 and 2009 makes sense.
Massive cash burn was an absolutely key feature of the dotcom boom/bust. Admittedly, it never really went away - there's always been a free->enshittification profit taking cycle since then. It's just the scale that's terrifyingly different this time.
They are counting on 'too big to fail'.
And when demand eventually crashes, all the new production capacity is left without buyers to sell to, so maybe it does not even make sense to create it.
There won't be any new production capacity until DDR6 comes out. There is no point in investing into an obsolete technology like DDR5.
China will happily take that business. They are producing DRAM now arent they?
Yes https://www.techpowerup.com/343185/chinese-cxmt-shows-homegr...
unfortunately not too promising: https://www.reuters.com/commentary/breakingviews/chinas-chip...
I wish they didn't stop producing DDR4, at least they'd be the sole producer of that.
You have to wonder whether they realy intend to use the ram, or they just spotted an opportunity to corner an essential market and choke/extort in resale.
I feel like what OpenAI has started to do was to accumulate as much compute resources as possible just so no one else has them.
> Right now it seems like these wafers will just be stockpiled in warehouses – like a kid who hides the toybox because they’re afraid nobody wants to play with them, and thus selfishly feels nobody but them should get the toys!
https://www.mooreslawisdead.com/post/sam-altman-s-dirty-dram...
I'm wondering this as well. Buying 40% of global production just sounds too much. What kind of user counts would they require for that much compute to pay off? Billions of people? What's the chance they could actually get that many users and charge them money? Zero?
Do they have the extra money to play games like that? I thought they were pulling back on ads because of the community reaction even though they wouldn't be introducing it now unless they needed it.
Money is like energy. You can flow it around a circuit if resistance is low.
They are not going to repeat the Windows 8 debacle again are they? (Over producing in response to perceived demand only to not get the demand when they are ready to sell) Didn't that bankrupt some memory manufacturers?
> Supply and demand.
Fake demand - fueled by ponzi debt financed by other debt.
Oh, forgot to mention - fake supply too. China was doubly-excluded from the RAM market by using both equipment export bans and tariffs - ensuring the supply is frozen solid.
I'm pretty sure this isn't a coincidence or due to incompetence.
To put this in long term relation.
Even current DDR4 3200 DIMM prices are at an all time high.
These are 6+ year old chip specs now!
I even thought stuff was overpriced four years ago in mid-2021 already, but this is a whole new level.
Some sample long term data for those:
https://geizhals.eu/?phist=2151624&age=9999
Between inflation generally, and DDR4 being obsolete and unsupported by current desktop or server CPUs, it would be unsurprising for DDR4-3200 DIMMs to be at an all-time high even without the current DRAM price shock. You can never count on old memory types dropping to bargain prices, because the major manufacturers are always eager to migrate the bulk of their production capacity to current-generation memory.
That process usually takes years.
Here the price hike was pretty instant as secondary effect of DDR5 evasion in two waves. I July and now in October.
There is usually no shortage of old working PC components as they also are avalable used and tested from people decommissioning and upgrading systems. These are not some rare parts in normal market situations.
I made a habit of maxing out motherboards a year or two before upgrading to an new platform. This was always dirt cheap until like 5 years ago.
Yeah the cheapest time to buy old tech is always just when the new stuff has come out. That's when suppliers are trying to shift old stock at cheaper margins.
You can take a look at the 5800X3D and how it was at its cheapest about 2 years ago when AMD was winding down production and Zen 4 had been launched.
To further your point, a used 5800X3D still goes for ~$300 when you can get a brand new 7800X3D for the same if not slightly cheaper (was on sale at MC for $280). I assume the high cost of DDR5 is pushing more people to not upgrade to AM5 and stay on AM4 for as long as possible. And most people are avoiding Intel chips out of principle after the chip degradation debacle - you can see that based on how much lower Intel motherboards are even though they have considerably better feature sets than the AMD equivalents.
Main reason for Intel boards being cheap is because they're practically one and done because of Intel's insistence on changing sockets every other generation.
Dang, I'm going to have to put my old RAM on ebay. I thought it was worthless, but I was clearly wrong.
Had CXMT not stop DDR4 manufacturing earlier this year, they'd at least cornered that market.
Unfortunately now they're too late on making the bank with DDR5, too: https://www.reuters.com/commentary/breakingviews/chinas-chip...
anyone want to buy a 2x16 GB DDR4-3200 kit that only fails memtest86 some of the time?
I wonder if semi-reliable RAM could be made to work for training. After all gradient descent already works in a stochastic environment, so maybe the noise from a few flipped bits doesn't matter too much.
Also, depends on the nature of the error. If only a small memory range is affected, you could patch the kernel to avoid it.
Have you tried underclocking or loosening the timings?
my experience is if even memtest86 fails the memory is truly borked.
Well, it kind of depends. With XMP (which is overclocking) I've found plenty of kits on Ryzen not passing memtest with the XMP settings. Different CPUs seem to be able to run their memory controller harder without error.
And then there are other factors like more sticks of ram stressing things further. I had to downclock to get memtest stable when running 4 sticks even though each kit ran fine on it's own. But that is expected as well as 4 sticks stresses the memory controller even further.
I confess I don't have any real recent experience with DDR5 though, mostly with DDR4 on Ryzen 1000-5000 series.
Sorry but I am looking for reliability. Do you have any stick that fails 100% of the time?
Now happy I bought that 256GB of DDR4 with the MB/processor combo of ebay last summer.
My 64 GB of DDR4 3200 RAM has tripled in price to $600. I am seriously considering selling it and just waiting for prices to drop again.
Maybe it's time we make a simple web page 100KB again? Is there some kind of CDN minification, adblocking and compression service? Maybe even server side rendering of websites?
Then a smartphone would work fine with 1GB of RAM and everyone could be happy.
Simple web pages can be made small with modern tooling, e.g. AstroJS.
The problem is that most web pages these days fundamentally are not simple.
Rather than trying to make web pages small, the real effort would be in designing web pages to be simple.
The large majority of software devs, PMs and the like don't really know how to do anything else than a Node + React webapp.
I've come to the opinion that for the vast majority of apps I've built, it could all be built using HTML + CSS (all built server side). I can sprinkle in little bits of interactivity using something like HTMX. And I'll have a website that is very easy to optimise, has phenomenal backwards compatibility, and gets rid of a whole class of issues associated with SPAs.
I often regret in my career not pushing back more on "requirements" that ended up requiring a more complicated app, whereas the customer would have been happier with a simpler solution.
I guess you're right, but it's more of a curve, though. Once you get to any decent level of complexity, it actually helps to have a framework instead of just going all HTML+CSS. Also it helps having something standard as react (that every web developer should fundamentally understand) than doing your custom stuff if other people will be working on it in the future.
There's a lot to say about the side effects of frameworks but there's a reason why everything converges towards that.
I think it's the other way around, a framework will get you up and running quickly, but then it becomes technical dept, and if your app is complicated you will end up fighting the framework. If you write something from scratch it will take a while to reach to the abstraction level where you can work fast. But then you have a fully custom abstraction layer that is not a "one size fits all" but custom tailored for your needs.
Good luck with hiring, onboarding, and maintenance of your bespoke solution. Also with your resume when seeking your next gig. For any serious project, ignore community and ecosystem health at your peril. To be clear, we're talking about framework selection, not leftpad vs DIY.
I agree, most websites allow complex flows. But I suspect, that most loads don't ever touch those. There is probably an automated way to deliver just a flat page, and maybe even allow for the top 5 interactions without loading all the frameworks and libraries.
And if someone could make a chat program not use 1 GB of RAM that would be great.
The majority of space is usually ads and I doubt websites want to reduce that.
Apple could make it happen. For some reason when an iPhone won't load something, people blame the thing instead of the phone. If they made Safari show an error page when a page used more than 256MB of RAM, suddenly the problem would disappear overnight.
No. Safari is already so buggy that it often crashes with the “a problem repeatedly occurred” message. That same message is shown when a webpage requests too much RAM. The problem did not disappear. People rightfully blame iPhone.
> 1GB of RAM and everyone could be happy
I remember a time when I was surfing the web with 256MB and a 5KB/s modem.
True, but I don't want to go back to that. Bloated web pages suck. But hi-res, true color images and videos don't suck.
Most web sites don't need that. Cat pics excluded, of course.
And I prefer text over video. Cat videos excluded, of course.
I was a bit shocked the other day when Chrome gave me a complaint about memory shortage and said I had something like 200GB in website data on my 256GB laptop(!) All put there by various web pages without my knowledge.
Obligatory mentions:
https://motherfuckingwebsite.com
http://bettermotherfuckingwebsite.com
Obligatory mention: https://www.downloadmoreram.com
Maybe it's time to not let a single company buy all the RAM.
Most people have phones that can handle webpages with 1-5MB JS bundles. Why artificially limit what you can do on the web? Why limit ourselves to 1GB RAM when more resources means tech becomes more useful?
Returning to simple webpages is popular idea on HN but it’s like wanting a car with no backup camera and crank windows. If your goal is to have your car be as simple as possible, then sure, but that’s not the case for most people.
Most people want their cars to be safe and convenient, and their webpages useful and rich, more so than they want to return to some idealized simplicity.
A simple webpage or blog with minimal styling that runs as an ARM binary on a TV remote is cool and fun but it’s not economically useful. It’s the equivalent of a manual scooter. We can build better apps (in the same way that car manufacturers can build less crappy infotainment systems) but optimizing for scarcity isn’t the answer in a world where abundance tends to grow.
(Edit: your downvotes mean nothing to me, I’ve seen what gets upvoted!)
I’d argue that in many of these instances, less is far more.
I want my car to just be really good at being a car, reliably get me from A to B. A Bluetooth connection to the stereo system is nice, but I don’t need a freaking 20” phablet right next to my face when I’m driving.
When I go to a website, I’m usually looking for information, to read something. I don’t often want fancy scroll and animations, I just want clear readable text free of distractions.
More and more these two examples seem to be going away, we’re losing the plot of what the point of these things are.
In a lot of ways, I agree with you. I think the key thing is that the complexity should be appropriate to what needs to get done.
Animations and etc. that distract from the actual content are superfluous. Agreed! I hate it when sites scrolljack.
But lots of HN posters want to impose the same austerity on every website, regardless of whether it’s appropriate. You can’t build Linear in 100KB of JS. Nor would you want to run it on 1 GB RAM. And that’s the case for a lot of economically useful applications.
Keeping things as simple as possible shouldn’t be the goal. It should be keeping it simple enough for the use case at hand.
Your mistake is assuming there is some correlation with usefulness and size.
The JS Gmail UI from 15 years ago was just as functional as the one today.
Websites that are supposed to be simple lists end up bloated and laggy because of really poor JS that makes one request per item iteratively to populate a list.
I do like the old JS Gmail UI. But the current JS Gmail UI doesn’t feel any slower. It is cluttered with more features, but some of them I find useful. (Displaying my calendar and being able to accept invites right in Gmail being a big one.)
As someone who used the HTML gmail interface right up until google pulled the plug: the JS version is much slower to load. Every morning, I get to have about 10 seconds thinking about how it used to be faster.
It absolutely is slower. To an extreme degree even. It takes 10 to 20 seconds to load and is incredibly sluggish to use on some low-end machines I use regularly.
I disagree that we should be optimizing for low end machines and holding back on product improvements for the 80% mass market. Technology improves, it’s one of its best traits. We don’t need to be stuck in the past.
Low-end machines aren't necessarily stuck in the past. e.g. $150 minipcs that use 5W to give you a perfectly snappy desktop Linux experience.
I think those machines are super fun and a snappy Linux experience is very satisfying. I use a lightweight WM myself on Linux and prefer it over the heavy ones. But this segment is 0.1% of technology users and we shouldn’t constrain applications to the limited hardware that this population chooses to use.
Actually a plain car would be great
I crank the window up and down 3x faster than the little button
And I could adjust my damn seat before electricity is available... sigh
Don’t get me started on automatic rear hatches. They close so slow it drives me crazy.
I can be in the house before it gets there, but I don’t trust. Just let me close it.
I’m glad you want that! But, most people wouldn’t. Also, electric seat adjustments give you way more options than the manual adjustments could. And typically with more precision than the under seat bar with discrete positions.
> But, most people wouldn't.
How do you know?
To me it’s so obvious that it doesn’t require testing. The market shows it; there are nearly no production vehicles at volume with crank windows.
Because the cost of another SKU exceeds the benefit, but you're saying the benefit is negative to everyone.
with you on manual seats? But crank windows? Nah man, power windows and locks are a requirement for me, as is a modern sound system, ac, and cruise control.
You can do a lot with little, it just requires investing more in development which understandably most companies are uninterested in. Besides, plenty of websites are bloated as all hell. Why does a newspaper website, for example, have to be very much more than plain html?
Newspaper websites are a good example of bloat, true. I think if you’re in the business of primarily serving text content and not doing much interactive stuff, you don’t need a heavy site. A lot of them tend to cram their websites with trackers and ads and I guess that’s a business thing.
Tbh, it’s unpopular around HN, but I felt like AMP was a great experience for users. AMP pages were super fast and had no annoying banners - and none of my pet peeve: layout shift.
I just checked my go-to stores in Europe. Holy shit. I bought a 5200-speed 16GB DDR5 kit in October for 55 eur. Now it's selling for 240 eur. The 6000-speed model is now 390 eur.
Starting to make Apple's RAM upsell prices look competitive...
I just checked our local apple store website and going from 48gb of ram on a M4 MacBook Pro it's 1500€ to get 128gb. Seems about the same :E
You actually get more, for the 1500€ you don't have to mess with sticks yourself. They solder it directly on your mainboard. If that isn't great service, I don't know what is.
I don’t know why you are getting downvoted. The sarcasm in the comment is pretty obvious.
I'm not sure if it's sarcasm or not, but either way it's just true
July: 172,38 € for a 2x32GB DDR5-6000 kit
Best price for the same one now: 689,55 €
I've been meaning to get more components for a diy NAS since atleast the last year and just been pushing it lazily. I'm literally kicking myself now when I actually started looking up deals for this black friday.
Same here, I was waiting for DDR5 to come down in prices to upgrade from 32 -> 128gb. Could have done it for 200-300€ a few weeks ago, now I'm just sad.
Depending on how many disks you need, buying old PCs on eBay is the way to go because there are still some Xeon E3 ECC models available cheap ($100) with plenty of RAM (16GB) and four SATA ports for a NAS.
Check used RAM. I've been reading those are still often fine, pricewise
The kit I bought last year for €200 is around €800 now. Crazy!
Been watching 2x32GB DDR5-6000 go from €360 to €920 on a local retailer over the last two weeks
They even have one SKU of 2x32GB DDR5-6400 that’s gone to… €4480
I bought a post lease mini PC with 16GB 4 months ago for the todays price of its RAM.
https://www.mooreslawisdead.com/post/sam-altman-s-dirty-dram...
The cynic in me thinks this would be a convenient way for these memory producers to manufacture demand, while also making OpenAI look good on paper. It’s not like they haven’t been caught price fixing in the past. Win win for these companies and a loss for everyone else.
> while also making OpenAI look good on paper
On paper this makes OpenAI look like absolute assholes. Like they have realized that all of their potential competitors will be memory constrained and have poured billions of dollars into making sure that happens instead of using that money to improve their own product.
If manufacturing fake demand (warehousing 900,000 memory wafers per month?) doubles and triples customer prices, it would decrease real demand, hurting the manufacturers after the artificial demand ends.
I think I explained that poorly. Basically artificially reducing supply so that these manufactures can get more for less. They've been caught doing it in the past before between each other, so why not use OpenAI as a bridge for that.
Fake demand? Either they’re selling the RAM or they’re not. They don’t make money by pretending to sell into fake demand. They make money by selling chips. A sale is a sale.
> Had Samsung known SK Hynix was about to commit a similar chunk of supply — or vice-versa — the pricing and terms would have likely been different
Wouldn't this be ... collusion?
Implicitly arguing that the memory oligopoly should have been coordinating is ... quite something.
OpenAI may well be doing something anticompetitive by cornering supply to foreclose competitors, but saying "they tricked the suppliers into not colluding!" is certainly a take you can have I guess.
No, it's not collusion to ask for more money from OpenAI if you hear that they are trying to buy 40% of the world's supply. Increased demand leads to higher prices, that's normal.
OpenAI, by doing simultaneous deals, hid the true demand from the suppliers, thus lowering their price and raising everyone else's.
A case could be made though that it is OpenAI who is operating illegally, attempting to corner the market.
There is nothing suspicious or abnormal about this behavior. It is called competition. Ironically, trying to prevent this kind of behavior prevents competitiob, and is a key factor for causing monopolization
It depends... if OpenAI bought the DRAM in order to use it, then fair play to them.
If they bought the DRAM in order to stop their competitors from using it because they are falling behind, that's anticompetitive in spirit, though I'm not sure if it actually breaks any laws.
I don't see what's less competitive about that, although I do see how it might make zero financial sense.
Knowingly attempting to buy or sell in quantities likely to move markets, for direct profit, is called manipulation and is most definitely illegal. this is true in physical markets commodity markets and financial markets. Not saying that this is what openAI is doing but it definitely merits an investigation.
It's illegal if you're trying to make prices higher, but it's completely legal and normal if you actually want the RAM.
Before you buy anything from a store, do you also show them all your receipts from the other stores you visited today?
> It's illegal if you're trying to make prices higher
That's exactly the goal, to raise prices for OpenAI's competitors.
How do you know they don't just want to buy that RAM to have the RAM?
All quantities bought or sold on the margin will move the market. Whether it moves the market is not up to the buyer or seller; it is up to other buyers and sellers who react to that transaction and adjust their expectations. This is normal market dynamics, dynamics we should want to happen because markets adjusting to movements of big players performs a social function; you and I need to know how large movenents of resources affects our livelihoods, and this is how that can happen.
There is no reason to pathologize or find suspicious these normal economic facts. Especially when it is not within the power of a big player to choose how other people react to their actions, which is all "moving markets" is. If something is suspicious and illegal about that, then it is equally suspicious that you and I seem to go along with this "market movement" by these big players and pay the new prices. Are we colluding with them? We could do with less conspiracy-minded interpretations of these things.
market economics are like newtonian mechanics. It's all so wonderful and logical and even elegant, until the dimensions expand a few orders of magnitude, and then all the rules break. Having worked on a trading floor for 20 years I know how this works. Swamping a market with huge trades is definitely considered manipulation by essentially all authorities, and indeed is a form of monopoly power, which even economic theorists will agree is undesirable. Jane Street just got a mega fine for exactly this in India, btw.
I can't get over the confident dismissal of science by hand-waving about imperfect modelling. But what I said has nothing to do with that, and is more true to the real world than an idealized perfect competition model. Pathologizing normal trading behavior like this is more the result laymen and authorities misinterpreting bad economic modelling. So I recommend you take some of your own medicine and look at the mirror. Maybe a trade affecting the market isn't so suspicious as you make it out to be, because the perfect competition model you're using to make accusations of monopoly simply doesn't make sense. Again, if there is something wrong with affecting the market, then you or I are just as liable for our consciously self-interested behavior of choosing higher-quality, lower-priced products.
What?? Replace DRAM with peaches or table legs and you have the exact example they give you during management training to explain implicit collusion.
That's when you raise prices without a change in the market conditions.
OpenAI is creating more demand, therefore the price must go up, if it didn't then there'd be shortages.
I can see how it can be confusing.
If you know demand will go up because Microsoft announced that each new Xbox will have 2TB of RAM, that is perfectly fine. Or if OpenAI issues a press release that they intend to buy half the worlds RAM.
If you know demand will go up because you learn the volume your customer intends to purchase from your competitor during confidential negotiations, that is not ok.
Every normal market interaction could be described by scary monopoly-sounding words. You heard about you're friend's salary in the same industry, and ask for a higher salary next time you negotiate. You're colluding with your friend to price gouge your employer!
In reality, so-called collusion is normal and unobjectionable. But when price surges happen, often due to factors outside of the seller's immediate control, people look for any reason to find an ethical dimension and find how to place blame, because this is more convenient. Things that were normal become abnormal and suspicious. It is in consumers economic self-interest to act in this way, because it often secures favors to them from various economic policies that they don't normally get when the market is "normal". This is no less a form of collusion than what sellers might do to secure their economic advantage. But a key difference for these anti "gouging" policies is that it gives consumers a special privilege and makes market pricing less able to fulfill its social functions.
They have, rather famously, been caught doing financial shenanigans before. [0]
[0]: https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
A competitive auction would not be collusion.
Only 50-60% of the recent price appreciation is after October 1st, so this is not likely the biggest direct cause, more likely it's everyone buying the finished stock on the market in a frenzy because they anticipate things like this will impact availability in the future, and it becomes a self-fulfilling prophecy.
Sounds like OpenAI might be trying their hand at TPUs, like what Google has. They are one of Google's biggest advantages in AI right now. It would also give them insurance against Nvidia being everybody's hardware supplier.
What do you mean “might be” trying their hand? It’s been widely reported that they are doing exactly that. It’s even known that they are doing it with Broadcom.
https://www.theregister.com/2025/10/13/openai_broadcom_deal/
This actually points the opposite direction, to doubling down on commercial GPUs.
NVIDIA recently told their board partners that they will need to source their own RAM and will not be bundling it with chips anymore.
If there is a supply crunch on DRAM, commercial GPU production lines will start having idle downtime. That is literally the worst possible thing that can happen to a company that has invested heavily in tooling and they will negotiate at or below cost production runs to fill the gaps if a customer can bring their own DRAM to the table.
Novice question: If they built something other than classic dram modules with the wafers maybe they could achieve faster bus speeds? How does Apple do it?
> How does Apple do it?
Apple uses off the shelf LPDDR modules. They have nothing whatsoever special about them.
Apple gets high bandwidth out of these modules not with high bus speeds, but with a very, very wide bus. This is expensive on the SoC side (requires a large die, which is why others don't necessarily do this), but allows for commodity memory modules.
That's what HBM is actually. The memory dies are directly next to the GPU die, on the same substrate. The main difference between Apple SoC and GPUs is the former use regular LPDDR while GPUs use HBM.
One of the key points of HBM is that dies are stacked up with many, MANY, more signals and channels. That's how NVIDIA has a memory bandwidth an order of magnitude higher than M4: 550GB/s for the M4 Max, 4.6TB/s for H200. And yes, that's bytes per second, not bits per second.
> while GPUs use HBM.
some GPUs use HBM. Most use GDDR. AMD and Nvidia still extract huge bandwidth from GDDR via high bus speeds + wide buses (like the 1.79 TB/s on the 5090)
Indeed! I implied "AI GPUs" since that’s where HBM is commonly used (despite AMD pioneering it on some consumer cards). And yeah, thousand-bit wide busses get close in performance.
shorter traces than soldered DIMM allow higher MT/s, this is fixed by CUDIMM/CAMM2, the other part of this is # of memory channels on the board, not sure why, but most consumer DDR5 boards have been 2 memory channels, you need to go to threadripper to get 4 or 8, It's unclear to me if this will still be an issue with future platforms.
You're not paying enough attention to the performance and cost impact of connectors and sockets. CAMM2/LPCAMM and CUDIMM have yet to be demonstrated operating at speeds that speeds that match the fastest soldered LPDDR, let alone GDDR; there's still a clear advantage for soldering memory.
CPU sockets with more than two memory channels are also far more expensive; the higher pin count usually increases the number of layers the motherboard needs, and the larger size of the socket requires more metal for stiffening (and EPYC CPUs still have issues with imperfect mounting leading to some IO lanes not working).
Using BGA soldering for both the processor and the memory sidesteps a bunch of engineering challenges.
> Using BGA soldering for both the processor and the memory sidesteps a bunch of engineering challenges.
by trading them with longevity challenges.
Even though 2 decades have past after the "usual suspect" lead-free solder, gpu or vram chips needing a reball is still a common occurrence from a cursory look at YouTube channels of professional electronics repairmen.
In case you needed more confirmation OpenAI has no IP moat...
That's blogspam. Actual Tom's Hardware article from October 1st:
https://www.tomshardware.com/pc-components/dram/openais-star...
That's not blogspam. There's a lot more in the mlid post than the linked news item above. Whether it's commentary or speculation I urge everyone to read and decide for themselves.
If these prices don't return back to normal I just don't see how Valves steam machine is less than $1,000 USD.
When Gaben's 11 organize a heist to steal the RAM wafers from the high-security vault in Sam Altman's basement, the price of the steam machine will drop below $1000 USD.
In before Valve releases a video of their own DRAM manufacturing plant because "they always wanted to have one"
They would already have contracts in place.
As someone currently dealing with pricing changes for DDR5 volume contracts (admittedly, only ~30k DIMMs, but still), those contracts are a lot less rigidly priced than we (engineering) realized at the time - “agreed price” is a lot more... flexible... until the pallet is delivered. Especially because our contract is with our manufacturer who has their own contract with a DIMM manufacturer who has their own contract with the DRAM manufacturer. DRAM is substantially more like a spot market than the average consumer would expect (at least at my volumes). The same is true for my HDDs (~100k unit/year contract) and my CPUs (10k unit/year contract). At least for HDDs our contract quantities are getting honoured (and we actually still have Seagate and WD fighting for our business), which I’ve heard hasn’t been the case for smaller orders.
Yep. It's a bit like some gas projects when prices spiked due to Russia's attack on Ukraine.
Companies had supply agreements in place but they will find essentially any excuse to "delay" delivery. You might eventually get your product but fat lot of good it does for you while you're mired in court and not getting anything anyway.
The particularly amusing example is a project continuously selling spot LNG while saying it's not yet operational, stiffing the companies with which they'd signed long term contracts with on a technicality.
For a while, yes. But I'm sure they will want to sell it for longer than a year or so.
Even Samsung is running into this issue now: their own internal foundry is refusing to give them a long term contract now so the S26 series will become more expensive.
If this happens to Samsung, what leverage will a player like Valve have?
> Even Samsung is running into this issue now [...]
These large corpos are so greedy to the point they harm themselves. I remember something similar with Amazon, where the Amazon shop had to redo the whole architecture from some microservice setup back to a monolithic approach because they used AWS and paid these weirdly structured prices like everyone else. Which made running a monolithic architecture under such constraints inherently cheaper. Not sure what the resulting compounded business cost for this "endeavor" was, but more often that not such things are never accounted for, so they don't show up as an issue.
That's their costs, not our costs. Regular PCs will cost more, which means Valve can price their consoles higher and collect margin, because there's no other competition/substitutes.
Maybe, but from a platform and brand development perspective, there may be smarter strategies.
Their interest isn't necessarily in squeezing out the most margin on each unit. If unexpected market conditions let them more or less hit their original margin targets but get far more units installed in homes, that could be much better for them in the long term.
There's a lot we can't know as outsiders, at the moment.
> that could be much better for them in the long term.
i agree. Valve's money machine is with the steam platform, rather than any hardware sales - breaking even on hardware is sufficient imho.
Valve's existential threat is from microsoft closing off windows somehow (or extracting rent...like a store!). Therefore, pushing steam machines which can be run without windows, is both business expansion as well as an insurance policy.
I'd wonder how much Valve has agreed to buy other parts and tooling/manfacturing services. Then they could have scenarios where they have a pile of AMD chips and the rest of the parts, but limited or no VRAM/RAM/NAND to complete the systems so they can be sold, or they've spent a pile of money designing and working with a manufacturer to get to the stage where they can start making batches, but can't get the volume to bring in a real benefit.
Being pragmatic, I wonder if Valve benefit from the health of the PC market, whether they'll also put efforts into helping people get more life out of their existing machines, or individuals/businesses refurbishing to sell on the used market with more confidence. There's already an angle for reducing e-waste with the win10 end of life if it can be tied into installing linux.
> i agree. Valve's money machine is with the steam platform, rather than any hardware sales - breaking even on hardware is sufficient imho.
Wasn't that true with the previous incarnation of the steam machine, the valve index, steam controller, etc.? IIRC their VR gear was some of the most expensive consumer level gear on the market.
Those controllers are first gen stuff - for the enthusiasts. They're expensive but not unduly so, not to mention that VR is still a niche, and is unlikely to expand the steam store. The first steam machine (steam box iirc) had both production issues and also not had much enthusiast success. Likely due to low hardware specs and not targeting the console market.
The steam deck, and now the (newer) steam machine is looking to push into the mass market of consoles, and i reckon it would be more successful. Not to mention the more mature steamOS and compatibility layer now (compared to before).
I have high hopes for it. I want PC gaming to break free of microsoft windows.
I upgraded my pc about this time last year (new CPU, RAM, and motherboard).
I was originally going to just get 64GB of DDR5-6000, with the option of adding another 64GB later, thinking the price might drop even further. At the last minute, I decided to get the whole 128GB instead. Glad I did.
I don't understand this. I'm looking at prices on some Asian store fronts and it's nowhere even remotely near these. I'm looking at DDR5-6000 2x16 for about $130, with no apparent limits.
Even with tariffs, transport, and other fees, you could get this to the US for way less than $400. I doubt the market could be this inefficient - in other words I don't think I just found a get rich quick scheme. So, what gives?
The current prices are a response to future stock, not current. It's 100% retailers price gouging with their current stock that they got for cheap because they know there will be limited stock in the near future. Asian retailers may be more honest and keep their margins the same, but will catch up in a month or two.
It's just surge pricing to manage demand, not "price gouging". Being able to buy RAM at high prices is a lot better than being unable to source it at all because everyone else is panic-hoarding.
> It's just surge pricing to manage demand, not "price gouging".
This carries the same energy as company leadership insisting that a RIF is not a layoff.
Price gouging has a specific meaning. It’s when costs are raised rapidly only because people are desperate and have no other choice. (e.g. water after a disaster)
Retailers raising prices in reaction to an incoming event that will take supply away from everyone is not that.
We don’t really know whether that’s true, since it’s hard to prove a negative (i.e., suppliers aren’t colluding). But given their history of price fixing it may be worth looking into.
Is it better? As with all price gouging, better for those who can afford it, sure, but not for those who can't. The proper way to combat scalping is to implement fair allocation methods (for a start purchase quantity limits) and punish people for scalping.
Look at how most places handled war-time gasoline shortages. Rationing coupons, purchase limits, demand leveling (like the odd-even system), price or profit controls, strict prosecution of scalpers and price gougers. And it's not like only the communists did this - even the US had most if not all of these things. And it worked far better than the shit that happened during the pandemic shortages. Governments used to know how to govern.
If you want to be "fair" for a necessity such as gasoline, you can have tradable rationing coupons. That way you are rewarded if you buy and use less gasoline, but the excess windfall due to the shortage is still transfered to you and away from the supplier. But even this assumes that gasoline is in fixed supply and there is no way of increasing its total production by paying more, which is not a very good assumption.
In a time of shortage, throwing more money at the problem usually won't increase supply. A shortage necessarily means that if you make more of something, you're guaranteed to sell it basically instantly, so there's already an incentive to increase production.
And it's not like the higher prices mean more money goes to the producers so they can invest in more production capacity. The price increase is spread out between every middleman in the chain untils there's almost nothing left. This could work only if the producers themselves are the ones raising prices, but then everyone else would still add their own cut, leading to even crazier price hikes, and also it's unlikely that extra profit would go to much more than lining the owners' pockets.
Additionally, demand spikes usually don't last, so any new production capacity you build will be a liability later, after the market settles down.
>A shortage necessarily means that if you make more of something, you're guaranteed to sell it basically instantly, so there's already an incentive to increase production.
This is patently false. Every oil reserve around the world has a cost per barrel of extraction. At $60/barrel many of them are shut down.
If you fix the price at $60 and demand goes up, you’ll end up with shortages and producers won’t be able to fill the gap.
It is very rare to have a market of physical goods where the cost of production is fixed and supply is effectively limited. For every other market, the price needs to go up to entice investing in making more supply.
The high prices ("price gouging") perform a social function because some cannot afford it; they prioritize what matters to them more, and ram is left available to those who absolutely require it. Trying to get around that by forcing prices below the market price simple encourages scalping behaviors. If prices are below the market price, but at the market price you would prefer the cash in your pocket over the ram stick, then you have every reason to sell it higher than you bought it because people are willing to buy it. It is those willing to buy it that are the main culprit in establishing the true price.
In reality, it is almost never a true binary of "afford" or "cannot afford" like critics of surge pricing make it out to be; people evaluate the price according to their circumstance and make a trade off. It is because of these decisions, the state of demand, that surge pricing is possible, not because of the machinations of evil price scalpers. That is why manufacturers couldn't lower prices even if they wanted to; gpu msrp being a great example of gpu vendors being caught between consumer ignorance about economics and the facts of reality that gpus are scarce enough to warrant higher prices.
I'm definitely not saying it's a binary can/can't afford situation. The point is that people don't "afford" things equally. I might some need RAM so badly that I'm prepared to take a huge risk and spend half of my paycheck on it. But that doesn't matter because someone else has 3x my paycheck, savings, and investment portfolio and a good credit score. How much money someone can spend on somethong is no indicator of how much they need/want it.
Something like "GPUs are actually scarce" doesn't even make sense to say, since scarcity is more a function of demand than supply. The supply of GPUs wasn't exhausted because people suddenly needed more GPUs or because Taiwan couldn't produce as many of them as they used to, it was because a few rich bastards were buying into a bubble so they could make as much money a possible before it all comes crashing down. They didn't "need" those GPUs much more than even the scalpers. They were just a vehicle to make short-term profits at the expense of everyone else.
And yes, of course those willing to buy things are the ones enabling the peice gouging. But that's not a useful observation. You either need something, so you'll buy it even if it doesn't make financial sense, or it makes financial sense to buy it, so you will. Notice how scalpers also fall into that second category, along with the rich bastards draining the supply.
Scarcity is a function of demand. That doesn't change the importance of what I said! In fact, it proves my point. The reason that "some rich bastard" is willing to pay such prices for gpus or ram or anything else is because they expect it to generate revenue for them equal to or greater than the amount they paid for it. How do they generate revenue? By selling a product! Who do they sell the product to? People like you and I who also compete with them to buy ram and gpus. In other words, the importance of what I said about prices depending on demand is that it depends on your demand. Rich bastards don't care much about ram per se; they could get it anyway. They care about it because of what (they expect) you're willing to pay for its products. Prices are what they are because people, on net, would rather spend the equivalent they could spend on ram by buying products of ram.
Putting this in view to the idea that people don't "afford" things equally: by your assumption, this implies people can indeed "afford" other non-ram things better when it comes to the more important alternatives they could buy with the ram-equivalent funds. Not only do ram-equivalent funds compete with alternate uses, but ram as a factor of production competes with other factors of production. And all factors of production compete, by way of the so-called rich bastards, for your and my dollars. In other words, if ram is more expensive, it is to support alternate uses of ram whose products are valued more highly by consumers than the direct use of ram. And, most importantly if one were to try to get around this higher resulting price for ram, it would cause higher prices for the products of those alternate uses of ram. People would be less able to get the thing they value more highly than ram because ram competes with all our needs, and less ram can be used for its indirect use.
All of that is to say that efforts to combat so-called price gouging bounds those who can less afford ram to be in a worse spot than otherwise. They can't afford ram as before, that much is true. But they prefer the alternatives to ram. If they would be better off by having ram, they would purchase that. Waving a magic wand to redistribute ram to them will give them ram, but now they lose what they valued more highly than ram.
Your logic that OpenAI can by proxy afford to buy out the entire world's supply of RAM because consumers value OpenAI products more than other RAM-dependent things assumes that OpenAI's money comes from selling goods or services to those consumers. It doesn't. The overwhelming majority of people pay zero dollars per year on AI services, while almost everyone spends at least a few hundred a year on gadgets that need RAM and other services that run on servers that need it as well.
The money OpenAI is using to starve the rest of us of RAM is coming from pumped up valuation through circular investments, investor FOMO, cheap debt and often straight up gambling. Rich bastards know that they can pump money into the bubble to grow it and hopefully cash out before it bursts. Nowhere in that process did any regular person value AI datacenters over other uses of RAM.
What you're pointing to is uncertainty; it is a judgement call whether or not consumers actually value these things. But if they don't open AI will lose money, assuming the market is not interfered with by bailouts, which would indeed hamper the social function of pricing. In a similar manner, this social function is hampered by interventions which manipulated credit and deby. My assumption only works in an economy where debt isn't artificially created ad hoc by the federal reserve and other banks. But that is a problem with how federal policies condition debt and finance in the modern economy, not with surge pricing. I would like to see more criticisms of the fed which essentially bankrolls bubbles like this rather than surging pricing, which performs its social function either way.
>since scarcity is more a function of demand than supply
This is incorrect. Anything that is bound by something like TSMC production is only made scarce when nvidia realizes they could sell out the entire run at $1000/card or whatever.
They were supply limited for years during the crypto boom. The way you know it was a supply problem is that you couldn’t even buy new cards because they were so frequently sold out.
Nvidia cards became really valuable overnight in the same way as any other asset. You trying to scream at “rich bastard” buyers will not change the fact that there is a shortage of cards so the price is going to go up across all sellers until the supply and demand curves intersect.
This is basic econ in action and history is fraught with attempts to try to fix supply shortages by capping the price.
> The high prices ("price gouging") perform a social function because some cannot afford it; they prioritize what matters to them more, and ram is left available to those who absolutely require it.
This is not true at all. It isn't left available to those who absolutely need it but to those who can pay for it. Those are two very different things.
You're attempting to construct a notion of need separate from the empirical experience of people's concrete choices, but that isn't possible if you want to be scientific. If a person spends $400 on a phone instead of 32gb of ram, they need the phone more than the ram. If a company spends $400 on ram instead of on some other production good, they need the ram more than the other production good - they expect it to generate more revenue.
These two exchanges are not disconnected either: phone prices are affected by revenues of companies which use ram for production. And those revenues are determined by purchases of phones. The person demonstrates through their choice that the use of ram for an indirect purpose of making phones (however indirectly that might effect it) is more valuable to them than the direct use in their computer. The person is not excluded from "having" the ram in the most general sense: they have it indirectly because they benefit from its use in production whose products they value more than the direct use of ram. The person, along with all other consumers, participates in organizing production in the manner that best benefits them, according to their needs, which may not necessarily involve them directly owning the thing.
During the pandemic there wasn’t gouging? I just remember complete shortages of inventory like toilet paper, basic microchips, etc.
It would have been better if people did raise prices during the pandemic for those things to prevent hoarding so I could actually wipe my ass at 2 cents a wipe instead of 1. But alas, the “price gouging” cry babies would have come out and lambasted them for “being greedy”.
Nope, governments were famously bad at this. Coffee rations and gas rations were a disaster.
(Edit: Replying here because of dumb rate limits)
>You guys are being unreasonable, we have plenty of toilet paper for everyone. Each person gets two rolls per week unless you can prove you need more until you calm the fuck down
And this is why it’s dumb. There actually was a supply shortage. You should read about it.
Toilet paper manufacturers made industrial scale toilet paper for offices, schools, public buildings, rest areas, etc.
1/3 of the entire toilet paper market for giant single ply rolls sold in bulk disappeared overnight. And that same demand flowed back into home multi-ply toilet paper that couldn’t be scaled up quickly because it came from a different mill.
Rations would have been completely stupid in reaction to a legitimate 50% increase in legitimate demand.
Read again. My whole point was that governments used to know how to handle a shortage, but they don't anymore, as evidenced by the pandemic. They let the market figure it out and you got a mix of both price gouging and scalpers.
And no, doubling prices wouldn't have done anything. Hoarders would just hoard more because not only was supply low, but prices were increasing, so they better buy it now rather than later.
If governments actually governed, this wouldn't have been a problem. "You guys are being unreasonable, we have plenty of toilet paper for everyone. Each person gets two rolls per week unless you can prove you need more until you calm the fuck down. We're also putting the toilet paper factory into overdrive to compensate for this stupidity."
It's speculative price gouging. Calling it "surge pricing" doesn't stop erosion of consumer trust in the market. Watch now as more people more readily jump to price fixing conclusions. Not helped by the inevitable further increase in speculation through feedback loops and the resultant volatility.
First come first served is a better principle than "surge pricing". A lottery is a better principle than "surge pricing". In the case that someone over purchased, they're free to dispose via secondary market if the value to them is lower than the out of stock price. I.e. decentralised pricing (and profits). Secondary market sales are just more efficient, they occur at negotiated prices that reflect true individual valuation, not the retailer's speculation.
I'd rather reward diligence and personal responsibility - if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price. Rather than passive reliance on wealth to solve problems. First come first served values effort and foresight. Scarcity is managed through time and effort rather than money.
> First come first served is a better principle than "surge pricing".
This is called a price ceiling, and it's a bad idea with a track record of failure and significant harm.
I'd rather pay extra and get what I need with 100% chance than get what I need cheaply with 5% chance and otherwise be forced to go without or buy from scalpers for the same price I would have paid anyways. This is the purpose of prices. So the people who really need it can buy it, and those who are borderline about the purchase decide to opt out.
If you're concerned with wealth inequality or one large buyer cornering the market, there are better ways to address those problem than prices ceilings.
> This is called a price ceiling
The act of eliminating surge pricing is not a price ceiling. That's a different thing. That requires more than simply swapping surge pricing with first come first served. You've created a strawman.
> I'd rather pay extra and get what I need with 100% chance
False dichotomy. Neither approach increases supply. Of course according to economists who can hand wave away bullwhip effects with simple "this model assumes X" statements that go unquestioned in the conversations which cite the findings of the given model but i digress. According to economists, both approaches do increase supply, the theory goes that the price gouging retailer invests in more factory capacity. Or the factory owner buoyed by vibrant secondary market activity views increased production investment as a safe bet. Maybe there's some truth in the latter...
> If you're concerned with wealth inequality
I'm concerned with lazy financial engineering over hard work. Why should the scrappy but innovative startup be excluded from resources over the sclerotic incumbent with a deeper wallet?
> The act of eliminating surge pricing is not a price ceiling. That's a different thing. That requires more than simply swapping surge pricing with first come first served. You've created a strawman.
What is it then? If you allow the price to increase, there is no need to enforce a different rationing mechanism. The only reason to think of implementing the alternate rationing mechanism is if the good isn't being rationed. If you insist on the previous, lower price, then that is a price ceiling in so many words, with the consequence of needing your rationing mechanism. If you allow a higher price, then your alternate mechanism is unnecessary.
The bullwhip effect is the whole reason why retailers may want to hold greater product stock in the first place; to absorb transient demand fluctuations and not have to pass them on in full to the supplier.
And a "scrappy but innovative" startup has an obvious interest in being able to source the DRAM or other goods they require, even at higher prices.
> The act of eliminating surge pricing is not a price ceiling.
If you force companies to price products lower than what they want, then you have a price ceiling by definition.
> Neither approach increases supply.
I didn't mention or allude to supply at all, although it's true that price ceilings also decrease supply (less so if there is monopoly or collusion, but they still do).
I was talking about resource allocation. The person who needs a new system because their previous computer broke will be willing to pay the extra money, but the person who already has a DDR4 system with a 5950x that runs their games well enough will be content to hold off on their AM5 upgrade to DDR5 because the marginal improvement isn't worth the extra $400.
If you have a price ceiling like what you proposed, the person with the DDR4 system may buy the DDR5 that they don't really need 1 day earlier than the person who actually needs the DDR5, creating a misallocation of resources.
(that's an example of the more abstract principle at play).
Theoretical arguments aren't even needed here. The empirical history of price ceilings is there for you to google.
If you didn't know that what you were proposing is a price ceiling, and you thought that I was talking about supply instead of resource allocation, then I mean this with no offence intended but you should study elementary economics before forming confident opinions on the subject. Society is in a vulnerable point with cost of living pressure and we don't need more energy behind these harmful populist ideas.
>> you have a price ceiling by definition
Price ceiling definition: a government-imposed legal maximum price
My original comment: First come first served is a better <snipped for brevity>
This is not a legal maximum price, this is a legal maximum in the derivative of price.
> I didn't mention or allude to supply at all
>> get what I need with 100% chance than get what I need cheaply with 5% chance
How do you square these two statements? One claims 100% supply certainty when no such thing exists in this context. Without making certain assumptions (unstated, but ludicrous, assumptions are rife in economics discourse), you can't state much of value about which buyer will get the goods in the surge pricing model, you especially cannot say that the buyer with the larger wallet will always win. Think for a second what assumptions you've made to this point in the conversation - you're still down the rabbit hole of price ceilings in the comment chain thus far.
>> The empirical history of price ceilings is there
Not disputed but as per your call out of definition above, not relevant.
To make the point further - the name for a limit on rate of change of price is not a price ceiling, anymore than the 0-60 time of a car is its top speed limit.
>> you thought that I was talking about supply instead of resource allocation
My challenge to you is to name the assumptions you've identified in your reasoning around resource allocation. I'm confident i can point out the deficiencies in your model because that is the nature of models.
>> you should study elementary economics
That's a great idea, a really good follow on from that is to identify logical fallacies you discover i. that process, especially those so accepted that it's not a stretch to say they are underpinning the discipline. A good example of that would be the conjectural origin theory of money but i digress.
> My original comment: First come first served is a better
Retailers are first come first serve. The first customer to buy product gets the next unit of inventory.
If you’re trying to argue that retailers need to be forced to set a retail price for each unit of in-stock inventory and then be forced to sell each unit at that price later no matter what the market rate does in the mean time, that’s an awful idea.
I’m sure you don’t mind when retailers decide to give discounts or put things on sale, do you? If the market price drops, they reduce selling price. Same thing when market rate goes up.
Forcing retailers to price things how you want doesn’t change supply and demand. It would only force retailers to add extra margin into their prices to account for the added risk of the government forcing their hand in sale prices.
> It's speculative price gouging ... if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price
The word for "monitoring market trends, anticipating needs and acting quickly" is, well, speculation. Why should a retailer not be allowed to speculate and hold more product stock when they anticipate a future crunch? In fact, the whole reason prices have become so volatile right now is that this supply crunch was not properly anticipated.
The reason why retailers are not "price fixing" is that price fixing involves setting an artificial ceiling on total production; retailers are not in a position to do this, and there is no evidence at all that DRAM makers are doing this either.
> First come first served is a better principle than "surge pricing"
First come first serve must means enterprising individuals would buy as much as they could afford and then immediately relist it on Facebook Marketplace for the actual market rate.
Thinking that retailers could keep RAM prices low and also keep it in stock for you is irrational.
>> Thinking that retailers could keep RAM prices low and also keep it in stock for you is irrational.
Not a claim i made
If your goal isn’t to make cheap RAM available to you, then what even is your goal?
Are you just angry that retailers are making a little extra temporary profit on their in-stock inventory? Is the goal to punish them for behaving rationally and understanding basic economics?
This perspective is pretty silly. If you don't know the price you set it up for auction. The Ebay auction prices in German are in the $400 range for 2x16GB. The retailers aren't particularly expensive. Maybe $10 to $20 more.
>I'd rather reward diligence and personal responsibility - if you monitor market trends, anticipate needs, and act quickly, such as buying RAM ahead of a known upcoming supply crunch, you're rewarded with access at the original price. Rather than passive reliance on wealth to solve problems. First come first served values effort and foresight. Scarcity is managed through time and effort rather than money.
Are you really going to discount the effort made by the manufacturers to produce the product? You know, giving them money is the best way for them to produce more products. A lucrative DRAM market will also make it easier for new competitors to enter the market.
Also, if everyone does what you suggest, then all that will happen is that the price will rise extremely rapidly before the actual supply crunch even happened. That sounds like what is currently happening.
> It's 100% retailers price gouging with their current stock that they got for cheap
The retail price of a product is a function of the market rate at the intersection of supply and demand. The price paid for inventory on the shelf doesn’t matter.
It works both ways. If retailers bought a lot of RAM at high prices and then the market suddenly dropped, they could have to sell it at a loss.
Some people get irrationally angry at this, but you do it too. If you bought a house for $500K and the market went up such that it was worth $700K, you wouldn’t think it was “price gouging” to list it at market rate. You’re just trading an asset for cash at the current price. The price you bought it at is irrelevant to the price you’re going to sell it for.
technically the house analogy isn't that good because there are lots of people who won't sell below what they bought for, even if there are no buyers at this price anymore. the house eventually sells if the seller is forced to sell, but price in such cases isn't necessarily 'correct' either. more liquid markets make better examples.
> technically the house analogy isn't that good because there are lots of people who won't sell below what they bought for
Just ignore the downside part of the house analogy because it doesn’t apply to this current DRAM situation where prices went up.
The key point is that if you buy something and its value goes up, you’re not going to offer it for sale at the same price you paid for it. You’re going to sell it at market rate.
Everyone does this, including retailers. It elicits angry cries of “gouging” from people who want to believe retailers are the cause of the high prices and be angry at them, but the market is big. It’s supply and demand.
Aus stock was cheap a couple of weeks ago, it's caught up though. I'd snatch it up if you need it.
Yeah, that's the spot market price, small sellers are more nimble, big platform sellers however probably feel the need to price in supply risks.
Shouldn't be recycled chips too, as those were always older gens. I bought one ddr3 dirt cheap and no issues.
The market is inefficient. Also, if you are in a small country, the demand is not there. So the seller will sell with regular prices. The stock will also be very low to export it back to where it came from.
Is it possible they are fakes?
As I mentioned in a prior post 7/18/2023 - G.SKILL Trident Z5 RGB Series 64GB (2 x 32GB) 288-Pin PC RAM DDR5 6000 - $253.
Today - G.SKILL Ripjaws S5 Series 64GB (2 x 32GB) 288-Pin PC RAM DDR5 5600 (PC5 44800) Desktop Memory Model F5-5600J3636D32GX2-RS5W - $620.
Prices from Newegg.
How to get every tech enthusiast in the world to root for OpenAI to implode in order to release the RAM hostages.
Not only that - I think OpenAI and all those other AI blobs must pay compensation damage to The People for inflating the prices here and skyrocketing costs.
Been planning to build my daughter a gaming PC for Xmas. Feel lucky I found a pre-built with 32GB RAM in a warehouse store on black Friday that hadn't been marked up yet. First pre-built desktop I've bought in decades and it was actually cheaper than I could build myself due to the crazy RAM prices.
If things go the way they are, I wonder how long until buying a used Playstation 4 (8GB DDR5 RAM) at around 100USD a piece to put Linux on it and use as a server in cluster will become viable thing to do? Bring back console clusters!
https://en.wikipedia.org/wiki/PlayStation_3_cluster
8GB GDDR5* Ram. Huge difference and one of the reasons the PS4 slapped the Xbox One around.
Or just remove the RAM and put it in more useful hardware.
Apple system config upgrades not looking so bad anymore
On this graph, the most expensive RAM is $12.5 / GB ($400 for DDR5-6000 2x16GB). Apple charges
- $25 / GB ($200 for 8 GB for the M5 MacBook Pro and the M4 MacBook Air)
- $16 / GB ($400 for 24 GB for the cheapest M4 Pro MacBook Pro)
- $12.5 / GB ($200 for 16 GB and then $800 for 64 GB more for the most expensive M4 Pro and M4 Max MacBook Pros)
and Apple's RAM is faster than PC RAM.
> Apple's RAM is faster than PC RAM
Are you sure about that? The M5 memory has a max bandwidth of ~150 GB/s, meanwhile there is PC memory that reaches 200 GB/s
m4 max is over 500
Came here looking for this comment.
I wonder how much price increase it takes for Apple to raise theirs.
Guess that’s the flip side of selling sometimes three to four years old hardware with zero price cut (I know they’ve been doing better in the M-series era)—the price also stays there when it’s skyrocketing elsewhere.
Exactly. Gotten from ‘price gouging’ to ‘not unreasonable’ in a couple months.
Some stores are going to a 'market price' system because the are tired of constantly printing new price tags.
https://www.pcworld.com/article/2984629/ram-is-so-expensive-...
What specifically happened in June to set this off? It can't just be "waves hands AI/LLMs" because ChatGPT has existed for years.
> What specifically happened in June to set this off?
Tariffs implemented by this administration:
0 - https://www.reuters.com/business/us-inflation-expected-rise-...The tariffs are bad, but you're simply wrong here - US tariffs do not affect global prices, it's OpenAI buying up production in advance and cornering the market.
There's a 4 month gap between June and October
Why did European prices have the same increase then?
> Why did European prices have the same increase then?
Where do Europeans get their DRAM from?
If it is the same handful of companies the US gets their DRAM from, then why would Europeans pay any less? Because the EU is not engaging in the same asinine trade war?
Sounds good in theory, but in practice those same few companies can set prices for markets outside the US to be at/near US prices. It doesn't take much effort for manufacturers to set their prices at or near those of their competitors and rely on an implicit mutually assured destruction[0] understanding.
0 - https://en.wikipedia.org/wiki/Mutual_assured_destruction
> Sounds good in theory, but in practice those same few companies can set prices for markets outside the US to be at/near US prices. It doesn't take much effort for manufacturers to set their prices at or near those of their competitors and rely on an implicit mutually assured destruction[0] understanding.
If a company in A sells a widget W to both the EU and the USA, such that a consumer in the EU and the USA pay the same prices even though the USA has a tariff and Europe doesn't, then the company will make a lot more profit per unit selling all their W in the EU and none in the USA.
I'm not at all sure what's happening at any given moment with the USA's tariffs on anything, given the chaos over there. But let's say W is the set of all things relevant to AI data centres. What this means is that all the data centres are now much cheaper to build in Europe rather than in the USA. Data centres can be put just about anywhere, given they're used over the internet anyway. This means that the companies selling W would have all the demand they want for W in the EU, so they could sell all of their supply of W in the EU, so they could get a higher profit margin on all of it.
I'm not sure how much DC investment money is going to which parts of the world, but I am sure that if all the suppliers stopped shipping to the USA because they could sell as much as they could make everywhere else in the world for more profit (and the same purchaser price after tariffs), I would have heard about it.
> If it is the same handful of companies the US gets their DRAM from, then why would Europeans pay any less?
... because tariffs are paid for by the buyer?
Importing memory from Korea to the US means the importer had to pay a tariff. Importing memory from Korea to Europe means the importer does not have to pay a tariff. The company selling the memory gets exactly the same amount of money in either case.
I'm sure if they didn't keep the prices somewhat similar, you would have a bunch of people in Europe selling RAM to Americans.
> I'm sure if they didn't keep the prices somewhat similar, you would have a bunch of people in Europe selling RAM to Americans.
I was just about to edit my response to the GP to say the same thing. Let's explore this hypothetical situation a bit further.
Suppose there was a DRAM manufacturer named "Acme DRAM" which decided to have a separate pricing schedule for the EU reflecting the lack of insane US tariffs.
Some enterprising entrepreneur in the EU would establish a company in the country having the least US tariffs and resell Acme DRAM to US companies. Surely this would make money hand-over-fist.
Problem is, the US DoJ does not look kindly on this kind of enterprise:
This would then put Acme DRAM in the crosshairs of an already vindictive and erratic US administration, likely to not only hammer the entrepreneur (see above) but to also include tariff ramifications for Acme DRAM as well.All of this risk in the pursuit of lower profit margins by definition.
0 - https://natlawreview.com/article/what-every-multinational-co...
> Some enterprising entrepreneur in the EU would establish a company in the country having the least US tariffs and resell Acme DRAM to US companies. Surely this would make money hand-over-fist.
Re-badging is a thing that some companies do actually do, despite what the DoJ says.
> Importers of goods into the U.S. should expect criminal enforcement to accelerate in the coming months and years.
I am absolutely not a lawyer, but wouldn't "importers" be the USA residents, not the EU businesses doing the exporting?
I was thinking this would look more like a a bunch of smaller operations selling on Ebay or platforms like that.
Huh? It's not the manufacturers paying the tariffs, its the importer? US tariffs do not affect the margins of the manufacturer.
> Huh? It's not the manufacturers paying the tariffs, its the importer? US tariffs do not affect the margins of the manufacturer.
This is a case of second-order effects[0].
See this post[1] for details.
0 - https://research.gatech.edu/blind-spot-big-decisions-why-sec...
1 - https://news.ycombinator.com/item?id=46144761
Afraid not. This has not happened across the board to all components.
> Afraid not. This has not happened across the board to all components.
Afraid so. Industry impact from tariffs does not require them to be applied "across the board to all components." See here[0] for more information.
With erratic massive tariff proclamations, counter-tariffs are to be expected. All it takes for companies to inflate prices is to either:
The net result is directly or indirectly, the tariffs implemented and/or threatened to be so are a significant contributor to electronic component costs.0 - https://www.tradecomplianceresourcehub.com/2025/12/03/trump-...
OpenAI bought up a big chunk of global DRAM production:
https://www.tomshardware.com/pc-components/dram/openais-star...
"Waves hands at AI/LLMs"
The problem is that training these models and using these models has required exponentially increasing amounts of memory.
ChatGPT has existed for years and in those years it's userbase and model size has increased tremendously. Not to mention the fact that a lot of competitors have sprung up in the wake of GPT. Including the likes of cloud based open model hosting services.
Perhaps some tariff shenanigans?
A possible answer as to why this is happening. https://thememoryguy.com/some-clarity-on-2025s-ddr4-price-su...
Scroll down. The same trend is happening for DDR5 prices. It's AI demand.
More specifically, it's OpenAI quietly contracting away 40% of global DRAM capacity, only to build wafers of memory and shove it in a warehouse.
https://www.mooreslawisdead.com/post/sam-altman-s-dirty-dram...
They're not the only player trying to buy more RAM. Every hyperscaler is also growing.
AWS, MS, Oracle, and Google demand is growing ~100% in 2026. Nvidia demand is growing ~50% in 2026. Supply isn't forecast to improve at all until H2 2026 and it's probably ~all spoken for through at least EOY 2026.
That's from June 2025. It's not relevant to current situation — prices have been rising since October 2025 due to agreements that became public in the same month.
Is he saying the supply of ddr4 went down because they’re switching to ddr5?
Really hoping this comes down again soonish. My desktop doesn’t have many years left :/
it's a commodity, it'll come down eventually, but soonish? depends completely on if/when AI bubble deflates and datacenter projects get cancelled.
12-24 months is my best guess. I bought some at inflated prices today at a retailer that failed to update to the latest market price by 15%. I'm not waiting that long only to be faced with a dilemma about waiting even longer for DDR6 to become reasonably priced.
GPU prices never went down after the crypto rush replaced GPUs with ASICs. I hope this does not repeat with DRAM and AI.
I picked up a pretty cheap second-hand GPU after the mining boom ended!
Somehow another sign that our world is collapsing.
We have more and more wars, less free trade and international spirit, freedom and privacy is going down the sink, "everything is going down", and now ram is a good example of things that use to be abordable and with price going down, becoming expensive luxury good.
One could say that RAM price is just a special temporary event not related to our down trend but a decade ago major players would not have left a profitable consumer market without new actors coming to fill in.
But now we prefer restricting offer instead of increasing production...
I think "collapsing" might be hyperbole.
It certainly seems clear that a purely upward trajectory of peace and living standards in the West was not supposed to last. History has its ebbs and flows, I'm sure we have a long way downhill to go, but I suspect there are peaceful times on the other side.
Net energy surplus had to end sometime, we have just decided to speed run it for some reason. Looking forward to sledding down the other side of the carbon pulse!
What could be the potential impact on smartphone and tablet prices in the coming months or years? I am assuming that laptop prices will start increasing next year.
Time for Intel to re-enter the memory market?
So, RAM outperforms stocks, gold, bitcoin, NFTs, real estate?
DDR5 is digital silver
I remember when crypto ruined the market for GPUs, and now AI is ruining the market for DRAM. Darn you, fads.
I wonder who the "visionaries" were that set the marked increase trend of max prices near 2024-12-09 for DDR4-3600 2x32GB and DDR4-3600 2x16GB?
I think that acted as a dog whistle for the rest of the market signalling there'd be no consequences for ripping other people off.
Also relevant: https://youtu.be/B7sB1-8jKno
I asked ChatGPT if it can generate a bunch of RAM for me. He says this is some kind of famous joke or something.
(I’m sorry I couldn’t resist)
"Downloading RAM" is more feasible than people think.
The seek time of a consumer-grade hard disk is said to be on the order of 10 ms. That's roughly the latency of a very high quality FTTH connection. Meaning that if you run a HDD rather than an SSD, a swap file in the cloud could potentially be faster than a local one (especially when you consider multiple reads/writes that could be done in parallel).
It's not exactly downloading more RAM but decently close to call it that for the joke.
Someone managed to boot a machine from a file system on Google Drive.
https://ersei.net/en/blog/fuse-root
Increasing swap is more like adding more HDD. It won't help memory pressure issues.
Yeah, not exactly just 1 million times slower.
It’s not clear to me how long the contracts last. The Tom’s Hardware article linked below mentioned “in 2025”, so, do we know if this monopolistic DRAM grab is nearly over or only just beginning?
Well, I guess one way these AI companies can stop us meeting our needs with local models - make it so we just cant get the hardware to run them...
The actual headline here is sort of misleading -- the 4x price increase is over the past 3 months!
I mean I wouldn't say it's misleading, it just says what the graph contains. It's important that it shows 18 months so you can see how flat it was before
I was planning to upgrade my PC 2 months ago, Corsair Dominator DDR5 96GB was ~ $400, and I thought it was a bit too much.
Well...
It hasn't been long since this happened The price hasn't been rising all year For d4 it's June, d5 is September so what happens
Maybe this is a "strategic inflection point" for Intel... to get back into the DRAM business?
They should get back into the Optane business first. Take the existing Optane memory to a modern PCIe 5 bus, and it will be competitive with lots of uses for DRAM. PCIe 5 gives you basically DDR4-like transfer bandwidth, and the Optane media has better latency and far better DWPD endurance than existing NAND SSD's. It's a total no-brainer.
Man, we can't even buy bare DRAM chips at work. Nobody will sell us anything. The whole US electronics industry has suddenly been put in a chokehold for.... perpetuating the AI bubble?
Man it'd be great if these AI assholes could stop absolutely trashing every single part of society, the economy, and the physical planet at large.
I hate this. PC gaming is my hobby, the only one that’s lasted my whole life. It’s always been there. It’s how I met my wife. It’s how I relax after a long day. It’s how I’ve participated in so many stories that stick with me and given me so many memories.
All of it is being murdered by the AI bros. Before them it was the crypto bros. It’s one thing after the other and I hate it so much.
Just live with the PC you have for two more years. It's probably not a big deal? A moderately capable machine from 5 years ago is still marginally capable.
Hopefully this blows over in a few years and you’re right and I’m just catastrophizing.
One way or another. Either the AI stuff cools off, or the RAM people spin up more fabs.
Well there's like 4 RAM manufacturers, so probably they will spin up new fabs but continue to collude to keep prices high.
They were previously colluding into the most profitable production volume. If demand stays high, the most profitable volume increases.
That means that yes, it will probably by more expensive than before this raise, but no, not nearly as much as today.
They already said they aren't going to increase supply, most likely because they think the AI bubble will pop and they'd be stuck holding the bag.
Businesses occasionally change their mind in response to market conditions.
Yh, my moderately capable machine is from 8 years ago and had already waited for an upgrade. Was going to be this Black Friday. Crabcakes.
There is more than a lifetime of incredibly great PC games that run on your existing hardware, and if this is your life's hobby, then paying an extra $100 or so every year or few is a drop in the bucket of your gaming expenses
I can afford it. Other people cannot, and the hobby is driven by a market existing for games. If newer people don’t enter the hobby as others die out, it fades away.
I mean its pretty rare to buy more RAM after completing your PC build + that single PC is going to last you 5+ years. Also mobos usually only have 4 slots in total so its not like its even going to take a lot. I'm rocking 2x48gb sticks and that's plenty for gaming.
The prices are wild tho.
I bought that ram in March 2024 for $384.81. Now it's priced at $1,172.99. LOL
It’s frustrating that it’s been one thing after the other, seemingly aimed directly at the thing I enjoy most.
The silver lining could be that game developers spend more time on actual gameplay, and less time on chasing graphical fidelity.
Silksong is playable on an 8 year old Nintendo Switch.
Yeah. Thanks for trying to keep me positive.
Hopefully this all calms down eventually. But it's hard not to feel like shit in this situation.
Increased demand for computer components for purposes other than gaming constitutes "AI bros murdering your lifelong hobby"?
PC gaming is not "murdered", it's doing better than ever.
In 2015 there were 3,000 games released to Steam, last year there were 18,000. In 2015 Steam's peak concurrent user count was 8.6 million. This year it's 41 million.
The inflation-adjusted price per gigabyte of RAM has dropped from $3/GB to $2/GB over the last 10 years, even including the recent price hikes.
So spare me the hysterics, your hobby is fine.
And you know what? The increased demand for compute always spurs innovation, so you'll probably get a better computer in the end as a result. You're welcome.
> In 2015 there were 3,000 games released to Steam, last year there were 18,000. In 2015 Steam's peak concurrent user count was 8.6 million. This year it's 41 million.
This is like saying "Spotify's subscriber count grew by 800% over the last 10 years. Music is doing better than ever!"
Who are the main players that make or benefit from these inflated DRAM prices?
Make? The market of course.
Benefit? The manufacturers of course.
what's funny for me, I bought this nice ram a while back but my CPU caps its speed so it's stuck at 2400mhz like ugh
I think most of us came to the conclusion that the AI hype is heavily responsible for this.
I want my money back. There should be an extra-tax on all those AI companies - they are heavily responsible for DRAM costing more now.
the other side of the AI bubble are a huge amount of 2nd hand parts of all kinds going to come onto the market?
Waiting for all my out of use 4, 8, and 16GB DDR3 DIMMs that I have somewhere in a drawer to become the new gold standard first.
I'm finding that getting an LGA socket to give me two reliable memory channels is far more valuable than the sticks that go into them. I've got at least two motherboards in use at my home right now with only one working channel of memory. There are sticks just sitting around.
My pile of worthless 512MB DDR1 sticks has bad news for you
Memory chips are divertied to registered server RAM and those will not fit most workstations
Nothing stops you building an EPYC based workstation. For the last socket you could even get mATX boards from Asrock Rack which were kind of ridiculous - I was very tempted to build my PC around one.
Unless it drags on for years so that the parts are old by the time everything gets liquidated.
Yeah though expecting that to A) take years still and B) the gear will be old AF
Bit like you could get NVIDIA Server cards before things went crazy but they’re on ancient cuda etc so not exactly as glorious as one would imagine
A lot of the goods won't be essily repurposible for consumer or small business work loads.
Imagine if auto manufacturers all refitted their factories and supply chains to produce military vehicles for a war effort. New family cars would run dry, and when the war ended, some folks would figure out make clever use of some surplus military vehicles for street travel and commerce, but most of the surplus would just be shifted to other military markets and family car production would take some time to resume.
AliExpress has been selling 2010 or whatever Intel xeons in dual socket desktop board kits for gaming. they are fairly affordable and hold up to almost modern games so people that can't afford new gen systems.
These products are fascinating. When I looked at them it seemed like they pulled old chipsets from server boards or maybe they found stock somewhere and then combined them with south bridges from things like Z77 boards. All to make use of obsolete server CPUs that were cheap and available. Amazing and wonderful.
Meanwhile you've got OpenAI buying up all the DRAM and maybe just piling it in a warehouse so no one else can have it or they figure out what to do with it. Microsoft recently said they don't even have a power source to plug the ton of GPUs they just bought into so they're also just sitting around collecting dust. What is even happening?
You can buy old rackmount servers on ebay for relatively cheap and used them as desktop PCs AFAIK (though they can be on the noisier side).
... and paying it back in power use
Use it during colder months for heating.
You don't want to be in the same room as these when they're running.
All in a pretty bad state of wear, I imagine.
At least with ECC memory it's very obvious when its failing as you'll see reported correctable errors.
I would snag up every possible H100 and H200 if the AI bubble burst and their prices went into steep decline.
Why? Doesn’t that suggest that demand justifies the “bubble”?
to make cheap and fast password hash cracking rigs
No, they'll be shipped off to developing nations to be dissolved for rare earths for the next boom cycle.
Memory chips don't contain rare earths, and e-waste recyclers pay a tiny fraction of what used-hardware buyers do.
My point was not to litigate the precise method of e-waste disposal.
Your point was to post nonsense without caring about whether it was true or false in order to troll someone into responding. Please don't do that again.
And your point was to nitpick a rhetorical statement and miss the point entirely. Feel free to disengage.
I don't think you should comment on this site again if you aren't willing to comment in good faith. I pointed out that your original statement was completely false, and rather than thanking me for the correction, you responded defensively and then attacked me. This shows that you don't care whether what you posted was true or not. That's corrosive to the presumption of good faith we employ here. If you aren't willing to behave properly, you shouldn't be here.
Stagflation!
Why is it that consumers always get shafted here and we just keep going on like "it's just business"
Why isn't some consumer protection regulator going "actually no, OpenAI, you can't corner the market for the foreseeable future"
Hell, make OpenAI pay for this shit up front. You want to corner the market? Put up the money you don't have.
Man, fuck these AI (and crypto) companies causing the rise of prices in GPUs and now memory.
At these prices, it might pay off to start an open-ish, non-profit, DRAM company to produce chips good enough for these purposes. They can sell a certain amount above cost for consumers and small businesses. Full price with guaranteed volume for some, enterprise suppliers.
The profits might be used to port the DRAM to multiple foundries to gradually increase supply. Alternatively, they can shift to produce other components, like VRAM. Make the low-to-midrange accelerators with larger VRAM more available at reasonable prices.
Just speculating. I don't know silicon economics well.
So it's doubled. From the complaints you'd think it was more like 10x.
On the flip side should the AI bubble burst won’t these be cheap if the data centres are liquidated?
Got real lucky. I was trying to negotiate a guy down on RAM. I wanted a server without as much RAM. In the end, I had to take the 768 GB. It only cost $1k more than 192 GB at the time. Unfortunately I can’t really sell off the excess because I need to substitute down the RAM modules to fill all 12. Ah well.
Moore's Law has been reversed. Instead of doubling transistors or capacity every 18 months, it is halved.
This is some kind of fundamentally different era for the tech industry, but ...
Inflation metrics seem to fail to capture the increase in leading-edge tech products of the past 18 months.
If we measure the value of the dollar against megabytes of RAM, well, in 01987, the difference between a StarBoard with "2 megs space" and a StarBoard with "2 megs installed" was US$484 (https://archive.org/details/amazing-computing-magazine-1987-...), so a "meg" of RAM cost US$242. Today https://pcpartpicker.com/trends/price/memory/ says "DDR4-3200 2x8GB" costs US$108; if that means 16GiB, as in 16384 "megs", a "meg" of RAM costs US$0.0066 today.
So the dollar's value has increased by roughly a factor of 36700 over those 38 years, averaging 32% per year.
That would be an average yearly inflation of -24%.
Too bad you can't live in DRAM or eat it when you're hungry.
> If we measure the value of the dollar against megabytes of RAM
We should be measuring it by the amount of RAM in a typical household PC in 1987 and today.
Even though a "meg" of RAM costs less than 1 cent today, I can't do anything useful with it.
Even if we are generous and buy a whole $1 of RAM today, it only gets us 150 MB of RAM, which, while infinitely more useful than 1MB, is still completely useless for running a modern OS/Browser.
What does your math say about that?
Math says that's stupid. You can still do everything with a "meg" of RAM that you could do in 01987. In fact, the "meg" of RAM you get now is enormously faster, and you can interface it to other things that you couldn't get at the time, such as SSDs, so you can do strictly more with it.
Economics says you're spending newly abundant resources freely in order to conserve those that are still scarce. Economics also predicts that people will adapt to RAM prices doubling by using RAM more frugally, spending more of other resources to compensate.
You can still run software from 1987 on a megabyte of ram, it just does different things from a modern browser.
They seem to fail to capture a whole lot of things. Supposedly $1 in 2000 is worth $1.88 in 2025. So 88% inflation over the 25 years. Meanwhile the median home price has increased by 150%. Family insurance by 350%. Median college tuition by 225%. Childcare costs have risen by 200%. But sure. We can buy super cheap 65" tvs now. Hurray for us! Literal kings who lived hundreds of years ago couldn't possibly imagine a world with cheap large screen tvs. So the poorest among us should rejoice at the wonders they are able to enjoy while they skip meals and ration their insulin.
A dozen eggs is up over 350%, but a 6-pack of Budweiser is only up 60ish% since 2000. So you know, it all balances out. Maybe drink a few extra cans of Bud with your next meal.
This explores the ideas behind your post: important things, like education and healthcare, have disproportionately risen in price while not-so-imortant thing have gotten less expensive.
https://www.yesigiveafig.com/p/part-1-my-life-is-a-lie
I don't exactly agree with the numbers, but I think the basic ideas are true
The inflation basket only represent a hypothetical average person, who doesn't even exist.
It's more useful to construct multiple separate inflation measures that represent different types of people. Like a "typical renter" inflation figure vs a "typical homeowner" inflation figure. It wouldn't be hard to do and would shine a light on inequality and help explain the rise of populism in certain segments of society.
An even better measure would somehow appropriately normalize the figure by the average disposable income in each of the segments to come up with a figure that measures the felt impact better.
The figure would be negative for wealthy people (who actually benefit from inflation because of asset price inflation) and positive for poor people (whose disposable income mostly goes to rent).
You can go one step further than that and calculate a fairness measure using something like the Gini coefficient (*) and analyze how much it has changed over time.
[*] https://wikipedia.org/wiki/Gini_coefficient
Making America Great Again, Beer and TV.
[flagged]
What a ridiculous statement. We have a massive number of people in the USA relying on financial support for purchasing food. Not every single American is fat, especially those who literally can’t afford to eat.
There are lots of Americans who go hungry, but mostly outside the USA. Starvation deaths inside the USA are almost entirely secondary to medical conditions that interfere with eating.
Generally there is a hidden starvation (I'm not sure the exact term), when one consumes just enough calories to survive a bit longer but almost no vitamins and minerals to live normal life.
Similar to polluted air/environment such deaths will hardly be backtracked to the true roots.
The most recent were deaths in conjunction with COVID. Was it due to COVID directly, indirectly or it was so bad that COVID could not really make it worse...
the reality I observe does not align with your Dickensonian version of it. food is laughably cheap in the US.
Sure how does it compare to similar low, middle, high end cpu prices? NVME drives?
They don’t capture anything going up because that would shame dear leader