> Budget brands normally buy older DRAM fabrication equipment from mega-producers like Samsung when Samsung upgrades their DRAM lines to the latest and greatest equipment. This allows the DRAM market to expand more than it would otherwise because it makes any upgrading of the fanciest production lines to still be additive change to the market. However, Korean memory firms have been terrified that reselling old equipment to China-adjacent OEMs might trigger U.S. retaliation…and so those machines have been sitting idle in warehouses since early spring.
This seems to almost be mentioned off-hand, but isn't this a really bad and un-free market, and a much bigger issue? Korean companies are afraid of doing business with Chinese companies because of the US, because of retaliation? This was not the "free and global market" I thought we were supposed to have at this point.
If production lines of DRAM are hindered by the politics of a unrelated 3rd party, then this seems to be a stronger cause of the current shortage than "a very large customer buying a lot in a short period of time".
> This was not the "free and global market" I thought we were supposed to have at this point.
Perhaps you haven't noticed but the pendulum has been swinging the other way for a while already and has a lot of momentum behind it. It's mentioned off-hand because the ongoing return to a multi-polar global order is covered elsewhere already, across dozens of articles every day.
Err… is there any question that the US is trying to slow down China’s high-tech computer development? I thought that was our open goal.
Countries decided the extent to which they’d like to engage in free trade together. It is a knob that we’d hope our leaders would turn strategically. (Regardless of whether or not we think our leaders actually are doing a good job of it…).
On paper it can sound rational. In reality you look at stuff like cars, for only so long people will tolerate buying a car for $60k when other countries, whom you are also competing with, get buy similar cars for $10-20k from China. Those same vehicles are used to boost productivity in your own domestic industries.
There is always a ton of risk involved with protectionism. Primarily whether your taxpayer-subsidized domestic jobs and hypothetical national security risk significantly outweighs all the very real economic costs.
We left behind any pretense of a free global market once we entered a post-tariff world. You can't have large universal tariffs or even the threat of them and expect the market to act freely, the two are fundamentally incompatible.
I don't think you can have a "free and global market" when countries participate in large-scale state industrial policy. Given those constraints, you have to either enforce a zero-subsidy environment (the US has no power to do this) or you have to accept that trade control is one arm of your foreign policy goals and surrendering it entirely is unlikely to help your aims.
For the most part, free and open trade is beneficial to the Western world order. But I think it's quite straightforward to imagine conditions under which it is not, many of which are currently in effect.
US control of EUV technology is probably the most obvious present one, but limitations on nuclear proliferation are an obvious case where there is no free market. Even selling civilian nuclear technology is controlled.
You may think of it analogously to Free Speech. The dream is complete and total expression. The reality is that if you allow convincing enough liars, your society starts to falter. Consequently, certain kinds of expression are not permitted - notably defamation. Think of it as more a North Star navigation ideal constrained by the trade winds (I suppose the Westerlies would be more relevant, but I couldn't resist the pun).
If you want a couple of reads, I enjoyed A Splendid Exchange about the history of trade, which I followed by the resurgent-though-once-dismissed Zeihan's Disunited Nations (which is more a hypothesis book than a history book).
It's becoming increasingly clear that OpenAI is going to get lapped by Google on technical merits. So this is the "code red" solution? Supply shenanigans?
They are getting beat in the developer market by Anthropic. And getting beat on fundamental tech by Google. This is a company whose ostensible mission is to "benefit all of humanity" ...
As in producers not over-producing RAM should be illegal? A presumably short-term price spike in RAM of all things is a non-issue. It is a luxury good that only a very small number of people care about and there is no reason to think this blip is going to last. Apple did stuff like this all the time at their high point in the late 2000s and early 2010s, and it would happen often in other markets. The world is not static and sometimes the situation changes and lots of supply is soaked up.
Who in the developed world doesn't have a few luxuries? Pretty much all of history people have had to make do with RAM being a lot less accessible than it is now. It isn't essential and people can still buy RAM in the rare situations where they actually need it.
There is nothing here worth invoking the legal system over. OpenAI can buy huge amounts of RAM if they want. Good luck to them, hope it works out, looks like an expensive and risky manoeuvre. And we're probably going to have a RAM glut in a few years looking at these prices.
DRAM is one of the categories of advanced semiconductors that the US considers important enough to national security that exporting it to China is forbidden. It's a fundamental industrial product.
Market manipulation is a crime under the Securities Exchange Act of 1934. You can't buy things to influence the price or the market, only to use or resell.
That'd probably make more sense if there wasn't also 50 other tech companies buying up RAM for the same reason (a sudden huge spike in demand due to AI taking off).
They mean to resell them in a different form: as part of their PaaS or SaaS. Per the article, OpenAI is just hoarding the wafers, not purchasing the final product.
It's about volume, not a naive count of consumers. Article claims that OpenAI holds contracts for 40% of world DRAM production. That's just really obviously manipulation if they can't actually power those chips, come on.
So the prosecution will gamble that OpenAI won't in fact use the RAM in a relevant timeframe and they only bought them to exclude the other swath of AI companies from competing?
From the article
> 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!
I guess we'll have to see if they in fact just keep "unfinished" RAM in warehouses like the article says and not roll them out into datacenters for a legitimate use as they are finished.
Should be, as in, new legislation should criminalize it? What's the generalized principle? Or should be, as in existing law should cover it? And if so, what law / how?
It wouldn't shock me if this is actually just market manipulation. OpenAI in the past year seems to be operating more and more like a pump and dump machine. Their recent AMD deal seems to have been AMD giving them a bunch of stock for free in exchange for them announcing that they would use AMD GPUs for training, and OpenAI doesn't have any fab equipment so the only thing they can do with 40% of the global dram supply is sell it to someone else.
I like how one of the reference links betrays how the article itself was researched, possibly written; HN hides the end of the url, which is "utm_source=chatgpt.com":
Guys, these are silicon wafers not bars of steel. You can't just stockpile 40% of annual capacity in a warehouse long-term. I would be incredibly surprised to find that any such large scale storage facility exists. Any storage of undiced wafers is temporary while the manufacturing pipeline proceeds. You've seen the pictures of clean rooms and stuff. There's no way you spend all that effort to make the wafer and then just stick it in a warehouse. Who even is going to make such a large cleanroom facility? And for what exactly? It doesn't even pass the basic sniff test.
Much more likely this is just a detail of the contract so that OpenAI can guarantee allocation. I would be surprised if the actual wafers entered OpenAI hands before being fully packaged.
There's nothing dirty about this deal. When making a large deal with one vendor he didn't disclose to them that he was making a deal with another vendor. That's pretty normal when you're trying to buy a lot of stuff. Otherwise, they can collude to shake you down.
I'm not thrilled about this genre of "guy I don't like does totally normal thing so it's bad". It's too engagement baity.
EDIT: Though even that may be wrong. TechCrunch reports that it was a joint meeting between the South Korean President, the heads of the two companies, and Sam Altman. I won't claim that TC is the bible but there's lots of stuff being reported that makes no sense, and this is a good deal for both these companies so it's more believable than news from someone that OpenAI is going to buy a bunch of wafers and stick it in a warehouse.
It's bad for consumers period. A deal that hampers 40% of global supply shouldn't be a thing, it's predatory. I know DRAM is not a necessity, but considering that PCs are going to be affected means this affects real things like schools and hospitals. There's being smart while making a deal and there's knee-capping the market with your leveraged to the tits business
> 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
It's not really that different from Apple reserving wafer starts on TSMC's next node and so on. It's just that this kind of capacity requirement has rarely shown up in DRAM before. Vendors prefer this kind of capacity reservation over a more variable finished product requirement. It allows them to know that they can build at the bottleneck rather than having to start up more capacity and then having that lie idle while everything downstream in DRAM packaging and DIMM production can't actually consume anything.
I don't buy it. It's easier to make this argument for companies that are building their own hardware, since they know it can be immediately used. OpenAI's move is tantamount to hoarding for the sake of strangling competition. There was plenty of supply to allow for their plans without this move (especially since they will probably go bankrupt at this rate).
It seems not normal (in the sense that it is obviously quite weird to but like half of the world’s RAM supply). But I wonder if they are also just not ready to announce what they are doing with it?
I mean with that many wafers, I guess it is possible that they’d be doing something pretty custom with the things…
> To be clear - 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
That's not "dirty." That's hiding your intentions from suppliers so they don't crank prices before you walk through their front door.
If you want to buy a cake, never let the baker know it's for a wedding.
What they mean is that they bought 40% of all RAM production, they managed to do that by simultaneously making two big deals at the same time. It's buying up 40% of all RAM production with the intention to have most of it idle in warehouses that is "dirty". And in order to be able to do that, they needed to be secretive and time two big deals at the same time.
That's not the dirty part. This is the dirty part:
> 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
> OpenAI isn’t even bothering to buy finished memory modules
And? Why should they be obligated to pay for all the middleman steps from fab down to module? That includes: wafer-level test, module-level test (DC, AC, parametric), packaging, post-packaging test, and module fabrication. There's nothing illegal or sketchy about saying, "give me the wafers, I'll take care of everything else myself."
> not even allocated to a specific DRAM standard yet
DRAM manufacturers design and fabricate chips to sell into a standardized, commodity market. There's no secret evolutionary step which occurs after the wafers are etched which turns chips into something which adheres to DDR4,5,6,7,8,9
> It’s not even clear if they have decided yet on how or when they will finish them into RAM sticks or HBM
The implication here is that the primary goal is to corner the market, not to use the supply. If you aren't going to use them anyways then of course it is silly to pay for them to be finished.
Do you think that's fine, or do you think that implication is wrong and OpenAI does actually plan to deploy 40% of the world's DRAM supply?
> The implication here is that the primary goal is to corner the market
You have no evidence of that. Even at face value, the idea of "cornering the market" on a depreciating asset with no long-term value isn't a war strategy, it's flushing money down the toilet. Moreover, there's a credible argument OpenAI wanted to secure capacity in an essential part of their upstream supply chain to ensure stable prices for themselves. That's not "cornering the market," either, it's securing stability for their own growth.
Apple used to buy-up almost all leading-edge semiconductor process capacity from TSMC. It wasn't to resell capacity to everyone else, it was to secure capacity for themselves (particularly for new product launches). Nvidia has been doing the same since the CUDA bubble took off (they have, in effect, two entire fabs worth of leading-edge production just for their GPUs/accelerators). Have they been "cornering" the deep sub-micron foundry market?
Yes, they've made insane scaling bets before and they have paid off.
If what we've heard about no acceptable pre-training runs from them in the last two years trying to increase the memory for training by two orders of magnitude is just a rehash of what got them from gpt2 to gpt3.
I'm curious how OpenAI has the funds to pay for 40% of the worlds ram production? Sure they are big and have a few billions but I kind of assumed that 40% for a year or whatever they are buying is easily double digit billions? That has to hurt even them, especially because they cant buy anything else?
Also what are these contracts? Surely Samsung could decide to cancel the contract by paying a large fee but is that fee truly so large that getting their ram back when prices are now 4x of what they used to be is not worth it?
I assume this includes more than just the raw price of modules but Openai only has 60 billion in funding altogether and was aiming for 20 billion ARR this year. This sounds like they are spending maybe half their money on RAM they never use? That just doesn't add up.
Secondary RAM Manufacturing Had Stalled. Budget brands normally buy older DRAM fabrication equipment from mega-producers like Samsung when Samsung upgrades their DRAM lines to the latest and greatest equipment. This allows the DRAM market to expand more than it would otherwise because it makes any upgrading of the fanciest production lines to still be additive change to the market. However, Korean memory firms have been terrified that reselling old equipment to China-adjacent OEMs might trigger U.S. retaliation…and so those machines have been sitting idle in warehouses since early spring.
My takeaway, this sounds like an comparably easy fix for the consumer market, if prices are somewhat guarenteed to stay mid term significantly above this years spring floor for someone to sweep up the margins and negotiate a somewhat reliable way to get the last gen production lines up and running again.
Will take at least half a year to pick up, but this is not a longterm RAM doomsday scenario in any sense.
I'm more worried about the low to mid-end embedded systems, that a have a dollar budget for memory components, that could get unbearably slow for the current/next gen if manufactures just use the bare minimum of RAM the bloated TV or tablet OS can run on, if the 1GB raspberry move is any indication of that. And consumers stuck with no way to upgrade them to a reasonably usable state.
One of the big problems here is that all of the hardware companies have been burned by hype before (e.g. crypto). No one actually believes that these AI companies will still be around in 5 years so spending billions to build factories for them doesn't make sense.
I don't know what companies will be around 5 years from now, but I would bet there will be more demand for RAM and the price per GB will be at least what it was before this price shock.
I remember it didn’t work out well for Randolph and Mortimer. Sam may pull it out, though, if he just sells the DRAM now while the market is still hot.
Yes, I have an older gaming PC from ~2018 that I keep putting off upgrading (first GPU prices skyrocketed, now this...) and was hoping to replace it with a Steam Machine next year. Will be endlessly bummed if that doesn't happen.
Will also be interesting if Sony/Microsoft was planning on releasing a next-gen system anytime soon, and I wonder if this will affect Apple's hardware at all.
They definitely can, as anybody could, for a short enough period of time. The only ones betting they'll be able to sustain it for a long period are the ones paying the currently very inflated price for ram.
And all these data centers they want to build around the country. When consumers can’t get devices they want maybe they’ll fight even harder against these data centers being built in their back yard. He’s not making any fans with this move that’s for sure
almost feel like OpenAI's recent "fall" is a decoy setup by them intentionally.. something's cooking.. maybe they wanted to buy back their own shares at a lower price?
OP is also marginally underestimating the impact this move would have on Google's competitiveness - they are making huge gains prototyping at light speed; this will halt their AI hardware acceleration plans pushing them back into slower software development on ever aging hardware.
It also shows why Nvidia is not afraid of competitors coming out with new desgings that obsolete their hardware: what good are superior designs with no fabs to produce them?
every one of these things that make the deal "good" for OpenAI is a direct result of negative externalities for everyone else: competitors, consumers, and people who wouldn't care otherwise.
The article even says that they don't have an obvious plan for how to use the wafers they bought, and very clearly suggests that this is purely an anticompetitive tactic to force everyone else to eat a price increase that OpenAI doesn't need to face. It's clever though because if any regulatory agency starts asking questions (not that they would do that in the current USA political climate) then OpenAI can just say it's a strategic reserve, we have plans to do something with it, etc. etc. What are you going to do? Take them to court and force them to auction off some % of the stock? Set an industry-wide limit on wafer inventory? Fine them? You'd need to find some evidence that it was done maliciously, and good luck with that.
There are some negative elements of captialism that we might simply have no reasonable regulatory apparatus to deal with. Preventing indivduals and companies from having so much market power in the first place seems to be the only thing that can work consistently.
As an AI researcher, I thought it was relatively well established (at least among my colleagues) that being pro-AI actually meant you were anti-Sam as well. He's the worst actor in the industry and has done an incredible amount of damage to its brand.
> Budget brands normally buy older DRAM fabrication equipment from mega-producers like Samsung when Samsung upgrades their DRAM lines to the latest and greatest equipment. This allows the DRAM market to expand more than it would otherwise because it makes any upgrading of the fanciest production lines to still be additive change to the market. However, Korean memory firms have been terrified that reselling old equipment to China-adjacent OEMs might trigger U.S. retaliation…and so those machines have been sitting idle in warehouses since early spring.
This seems to almost be mentioned off-hand, but isn't this a really bad and un-free market, and a much bigger issue? Korean companies are afraid of doing business with Chinese companies because of the US, because of retaliation? This was not the "free and global market" I thought we were supposed to have at this point.
If production lines of DRAM are hindered by the politics of a unrelated 3rd party, then this seems to be a stronger cause of the current shortage than "a very large customer buying a lot in a short period of time".
> This was not the "free and global market" I thought we were supposed to have at this point.
Perhaps you haven't noticed but the pendulum has been swinging the other way for a while already and has a lot of momentum behind it. It's mentioned off-hand because the ongoing return to a multi-polar global order is covered elsewhere already, across dozens of articles every day.
Err… is there any question that the US is trying to slow down China’s high-tech computer development? I thought that was our open goal.
Countries decided the extent to which they’d like to engage in free trade together. It is a knob that we’d hope our leaders would turn strategically. (Regardless of whether or not we think our leaders actually are doing a good job of it…).
> I thought that was our open goal.
Is the goal also to hurt South Korean businesses and all businesses in the world, just to "pwn China" basically?
We’re probably also spurring China to develop more independently. I don’t think it is a good plan, just an unconfusing one.
On paper it can sound rational. In reality you look at stuff like cars, for only so long people will tolerate buying a car for $60k when other countries, whom you are also competing with, get buy similar cars for $10-20k from China. Those same vehicles are used to boost productivity in your own domestic industries.
There is always a ton of risk involved with protectionism. Primarily whether your taxpayer-subsidized domestic jobs and hypothetical national security risk significantly outweighs all the very real economic costs.
"America First" as an ideology means that question is never considered
We left behind any pretense of a free global market once we entered a post-tariff world. You can't have large universal tariffs or even the threat of them and expect the market to act freely, the two are fundamentally incompatible.
Oh, the market will find a way around this too. The more US uses this particular button the less effective it becomes.
What free market?
I don't think you can have a "free and global market" when countries participate in large-scale state industrial policy. Given those constraints, you have to either enforce a zero-subsidy environment (the US has no power to do this) or you have to accept that trade control is one arm of your foreign policy goals and surrendering it entirely is unlikely to help your aims.
For the most part, free and open trade is beneficial to the Western world order. But I think it's quite straightforward to imagine conditions under which it is not, many of which are currently in effect.
US control of EUV technology is probably the most obvious present one, but limitations on nuclear proliferation are an obvious case where there is no free market. Even selling civilian nuclear technology is controlled.
You may think of it analogously to Free Speech. The dream is complete and total expression. The reality is that if you allow convincing enough liars, your society starts to falter. Consequently, certain kinds of expression are not permitted - notably defamation. Think of it as more a North Star navigation ideal constrained by the trade winds (I suppose the Westerlies would be more relevant, but I couldn't resist the pun).
If you want a couple of reads, I enjoyed A Splendid Exchange about the history of trade, which I followed by the resurgent-though-once-dismissed Zeihan's Disunited Nations (which is more a hypothesis book than a history book).
Moves like this should be illegal.
It's becoming increasingly clear that OpenAI is going to get lapped by Google on technical merits. So this is the "code red" solution? Supply shenanigans?
They are getting beat in the developer market by Anthropic. And getting beat on fundamental tech by Google. This is a company whose ostensible mission is to "benefit all of humanity" ...
As in producers not over-producing RAM should be illegal? A presumably short-term price spike in RAM of all things is a non-issue. It is a luxury good that only a very small number of people care about and there is no reason to think this blip is going to last. Apple did stuff like this all the time at their high point in the late 2000s and early 2010s, and it would happen often in other markets. The world is not static and sometimes the situation changes and lots of supply is soaked up.
Who in developed countries doesn't buy computers and by extension ram
Who in the developed world doesn't have a few luxuries? Pretty much all of history people have had to make do with RAM being a lot less accessible than it is now. It isn't essential and people can still buy RAM in the rare situations where they actually need it.
There is nothing here worth invoking the legal system over. OpenAI can buy huge amounts of RAM if they want. Good luck to them, hope it works out, looks like an expensive and risky manoeuvre. And we're probably going to have a RAM glut in a few years looking at these prices.
DRAM is one of the categories of advanced semiconductors that the US considers important enough to national security that exporting it to China is forbidden. It's a fundamental industrial product.
I don't follow how computers are not essential.
Seems like it is, but the question is whether the current Justice Department will do anything about it.
I read "US Justice Department" the same way I read "Britain's Ministry of Truth".
When I hear about this US justice department, I hear the mafia enforcement.
> Seems like it is
Do you have a citation for what law is being violated? Or just vibes?
https://fortune.com/2025/11/23/ai-rivals-like-openai-nvidia-...
Obviously kind of a moot point because whether it violates antitrust law or not, what is guaranteed is the US Government is not going to do anything
Whether the Clayton and Sherman acts apply to the Stargate initiative is not relevant to this RAM-hoarding activity.
Market manipulation is a crime under the Securities Exchange Act of 1934. You can't buy things to influence the price or the market, only to use or resell.
https://en.wikipedia.org/wiki/Market_manipulation
The law you refer to applies only to markets for securities. RAM is very clearly not a security, it fails the Howey test.
There are similar laws prohibiting the manipulation of commodity markets but I do not believe a US court would find RAM to be a commodity.
How is RAM not a commodity?
Do you think OpenAI plans to trade the semiconductor market? This would only apply in that scenario.
No, they want DRAM to be expensive to give them a competitive advantage over their competitors.
That'd probably make more sense if there wasn't also 50 other tech companies buying up RAM for the same reason (a sudden huge spike in demand due to AI taking off).
They mean to resell them in a different form: as part of their PaaS or SaaS. Per the article, OpenAI is just hoarding the wafers, not purchasing the final product.
It's about volume, not a naive count of consumers. Article claims that OpenAI holds contracts for 40% of world DRAM production. That's just really obviously manipulation if they can't actually power those chips, come on.
So the prosecution will gamble that OpenAI won't in fact use the RAM in a relevant timeframe and they only bought them to exclude the other swath of AI companies from competing?
From the article
> 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!
I guess we'll have to see if they in fact just keep "unfinished" RAM in warehouses like the article says and not roll them out into datacenters for a legitimate use as they are finished.
The current Justice Department? You're kidding, right?
> Moves like this should be illegal.
Should be, as in, new legislation should criminalize it? What's the generalized principle? Or should be, as in existing law should cover it? And if so, what law / how?
It wouldn't shock me if this is actually just market manipulation. OpenAI in the past year seems to be operating more and more like a pump and dump machine. Their recent AMD deal seems to have been AMD giving them a bunch of stock for free in exchange for them announcing that they would use AMD GPUs for training, and OpenAI doesn't have any fab equipment so the only thing they can do with 40% of the global dram supply is sell it to someone else.
the only thing they can do with 40% of the global dram supply is sell it to someone else.
The way it works is that OpenAI will have the DRAM delivered to Nvidia/AMD/Broadcom to be assembled into the racks that OpenAI buys.
china does this all the time fyi
An American company, combined with American tariffs, and fear of American retaliation.
Getting pretty tired of that place tbh.
I like how one of the reference links betrays how the article itself was researched, possibly written; HN hides the end of the url, which is "utm_source=chatgpt.com":
> https://www.economist.com/business/2025/11/19/cracks-are-app...
You have to appreciate the irony :)
If OpenAI were actually using the RAM that’s one thing - but stockpiling raw wafers in warehouses is egregious.
Guys, these are silicon wafers not bars of steel. You can't just stockpile 40% of annual capacity in a warehouse long-term. I would be incredibly surprised to find that any such large scale storage facility exists. Any storage of undiced wafers is temporary while the manufacturing pipeline proceeds. You've seen the pictures of clean rooms and stuff. There's no way you spend all that effort to make the wafer and then just stick it in a warehouse. Who even is going to make such a large cleanroom facility? And for what exactly? It doesn't even pass the basic sniff test.
Much more likely this is just a detail of the contract so that OpenAI can guarantee allocation. I would be surprised if the actual wafers entered OpenAI hands before being fully packaged.
Maybe they’ll build nice little forts with the wafers
If they do, it'll be a house made of glass.
Yup. Not exactly a move "for the good of the world".
Also consider:
Warehouses of small, high-value items that are fungible and untraceable.
That will create multiple huge targets for a big heist. And they'd best have good eyes on their security people too.
Sounds like something big enough for organized crime to target.
There's nothing dirty about this deal. When making a large deal with one vendor he didn't disclose to them that he was making a deal with another vendor. That's pretty normal when you're trying to buy a lot of stuff. Otherwise, they can collude to shake you down.
I'm not thrilled about this genre of "guy I don't like does totally normal thing so it's bad". It's too engagement baity.
EDIT: Though even that may be wrong. TechCrunch reports that it was a joint meeting between the South Korean President, the heads of the two companies, and Sam Altman. I won't claim that TC is the bible but there's lots of stuff being reported that makes no sense, and this is a good deal for both these companies so it's more believable than news from someone that OpenAI is going to buy a bunch of wafers and stick it in a warehouse.
https://techcrunch.com/2025/10/01/openai-ropes-in-samsung-sk...
It's bad for consumers period. A deal that hampers 40% of global supply shouldn't be a thing, it's predatory. I know DRAM is not a necessity, but considering that PCs are going to be affected means this affects real things like schools and hospitals. There's being smart while making a deal and there's knee-capping the market with your leveraged to the tits business
> 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
That is not "totally normal".
It's not really that different from Apple reserving wafer starts on TSMC's next node and so on. It's just that this kind of capacity requirement has rarely shown up in DRAM before. Vendors prefer this kind of capacity reservation over a more variable finished product requirement. It allows them to know that they can build at the bottleneck rather than having to start up more capacity and then having that lie idle while everything downstream in DRAM packaging and DIMM production can't actually consume anything.
Did Apple reserve 40% of the supply and cause a massive shortage. If not, then I don't see how its the same.
I don't buy it. It's easier to make this argument for companies that are building their own hardware, since they know it can be immediately used. OpenAI's move is tantamount to hoarding for the sake of strangling competition. There was plenty of supply to allow for their plans without this move (especially since they will probably go bankrupt at this rate).
How exactly could they 'hoard' this? There's no place in the world to store that much undiced wafer. It will all go bad.
It seems not normal (in the sense that it is obviously quite weird to but like half of the world’s RAM supply). But I wonder if they are also just not ready to announce what they are doing with it?
I mean with that many wafers, I guess it is possible that they’d be doing something pretty custom with the things…
> To be clear - 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
That's not "dirty." That's hiding your intentions from suppliers so they don't crank prices before you walk through their front door.
If you want to buy a cake, never let the baker know it's for a wedding.
What they mean is that they bought 40% of all RAM production, they managed to do that by simultaneously making two big deals at the same time. It's buying up 40% of all RAM production with the intention to have most of it idle in warehouses that is "dirty". And in order to be able to do that, they needed to be secretive and time two big deals at the same time.
> It's buying up 40% of all RAM production with the intention to have most of it idle in warehouses
They have no incentive to purchase a rapidly-depreciating asset and then immediately shelve it, none
They might have to warehouse inventory until they can spin-up module-manufacturing capacity, but that's just getting their ducks in a row
That's not the dirty part. This is the dirty part:
> 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
> OpenAI isn’t even bothering to buy finished memory modules
And? Why should they be obligated to pay for all the middleman steps from fab down to module? That includes: wafer-level test, module-level test (DC, AC, parametric), packaging, post-packaging test, and module fabrication. There's nothing illegal or sketchy about saying, "give me the wafers, I'll take care of everything else myself."
> not even allocated to a specific DRAM standard yet
DRAM manufacturers design and fabricate chips to sell into a standardized, commodity market. There's no secret evolutionary step which occurs after the wafers are etched which turns chips into something which adheres to DDR4,5,6,7,8,9
> It’s not even clear if they have decided yet on how or when they will finish them into RAM sticks or HBM
Who cares?
The implication here is that the primary goal is to corner the market, not to use the supply. If you aren't going to use them anyways then of course it is silly to pay for them to be finished.
Do you think that's fine, or do you think that implication is wrong and OpenAI does actually plan to deploy 40% of the world's DRAM supply?
> The implication here is that the primary goal is to corner the market
You have no evidence of that. Even at face value, the idea of "cornering the market" on a depreciating asset with no long-term value isn't a war strategy, it's flushing money down the toilet. Moreover, there's a credible argument OpenAI wanted to secure capacity in an essential part of their upstream supply chain to ensure stable prices for themselves. That's not "cornering the market," either, it's securing stability for their own growth.
Apple used to buy-up almost all leading-edge semiconductor process capacity from TSMC. It wasn't to resell capacity to everyone else, it was to secure capacity for themselves (particularly for new product launches). Nvidia has been doing the same since the CUDA bubble took off (they have, in effect, two entire fabs worth of leading-edge production just for their GPUs/accelerators). Have they been "cornering" the deep sub-micron foundry market?
Yes, they've made insane scaling bets before and they have paid off.
If what we've heard about no acceptable pre-training runs from them in the last two years trying to increase the memory for training by two orders of magnitude is just a rehash of what got them from gpt2 to gpt3.
"...their deals are unprecedentedly only for raw wafers — uncut, unfinished, and not even allocated to a specific DRAM standard..."
wtf. life sucks.
That $7 trillion number makes more sense now.
I'm curious how OpenAI has the funds to pay for 40% of the worlds ram production? Sure they are big and have a few billions but I kind of assumed that 40% for a year or whatever they are buying is easily double digit billions? That has to hurt even them, especially because they cant buy anything else?
Also what are these contracts? Surely Samsung could decide to cancel the contract by paying a large fee but is that fee truly so large that getting their ram back when prices are now 4x of what they used to be is not worth it?
I found this which claims ram market in 2024 was almost 100 billion: https://www.grandviewresearch.com/industry-analysis/random-a...
I assume this includes more than just the raw price of modules but Openai only has 60 billion in funding altogether and was aiming for 20 billion ARR this year. This sounds like they are spending maybe half their money on RAM they never use? That just doesn't add up.
I have a 32 GiB DDR5 set, happy to exchange for $500K in cash or a nice little house in Spain.
I have a nice little house in Spain, I'm willing to trade it for 128GB of RDIMM DDR5.
No issue pal, I have that too. I can survive on my 32 GiB set for a while.
4800MHz single rank ok?
Secondary RAM Manufacturing Had Stalled. Budget brands normally buy older DRAM fabrication equipment from mega-producers like Samsung when Samsung upgrades their DRAM lines to the latest and greatest equipment. This allows the DRAM market to expand more than it would otherwise because it makes any upgrading of the fanciest production lines to still be additive change to the market. However, Korean memory firms have been terrified that reselling old equipment to China-adjacent OEMs might trigger U.S. retaliation…and so those machines have been sitting idle in warehouses since early spring.
My takeaway, this sounds like an comparably easy fix for the consumer market, if prices are somewhat guarenteed to stay mid term significantly above this years spring floor for someone to sweep up the margins and negotiate a somewhat reliable way to get the last gen production lines up and running again. Will take at least half a year to pick up, but this is not a longterm RAM doomsday scenario in any sense.
I'm more worried about the low to mid-end embedded systems, that a have a dollar budget for memory components, that could get unbearably slow for the current/next gen if manufactures just use the bare minimum of RAM the bloated TV or tablet OS can run on, if the 1GB raspberry move is any indication of that. And consumers stuck with no way to upgrade them to a reasonably usable state.
One of the big problems here is that all of the hardware companies have been burned by hype before (e.g. crypto). No one actually believes that these AI companies will still be around in 5 years so spending billions to build factories for them doesn't make sense.
I don't know what companies will be around 5 years from now, but I would bet there will be more demand for RAM and the price per GB will be at least what it was before this price shock.
Guess OpenAI finally found a business model that works: memory futures.
I'm reminded of the 1983 deal to corner the market on Frozen Concentrated Orange Juice.
Or the Hunt brothers and silver which was just a few years before that.
How'd that turn out? https://en.wikipedia.org/wiki/Silver_Thursday#:~:text=On%20J...
That's a great comparison. The consequences are pretty universal too. History implies this won't end well for OpenAI.
I remember it didn’t work out well for Randolph and Mortimer. Sam may pull it out, though, if he just sells the DRAM now while the market is still hot.
"Mortimer ... we're back!"
Sell! Sell! Get back in there and sell!
I wonder if this kills Valve's Steam Machine and Steam Frame
I've been selfishly wondering the same thing. The Frame is on my shortlist (as long as the price wasn't too crazy).
Yes, I have an older gaming PC from ~2018 that I keep putting off upgrading (first GPU prices skyrocketed, now this...) and was hoping to replace it with a Steam Machine next year. Will be endlessly bummed if that doesn't happen.
Will also be interesting if Sony/Microsoft was planning on releasing a next-gen system anytime soon, and I wonder if this will affect Apple's hardware at all.
As is mentioned in the article, depends on when they bought their DRAM contracts. If they were in before this then they'll be fine for a while.
Why would one sell something cheaper than the current market value? I wouldnt care if i had stock or not, prices should be what things cost.
When someone buys up all the supply, the price will rise because of supply and demand. It's the nature of markets.
TFA mentions that if Samsung and SK Hynix had known what shenanigans were underway, they would have pursued better pricing terms.
Can OpenAI hurry up and go bankrupt?
Force a divestiture of Microsoft.
nice 5d chess move, saltman paobably got this idea from GPT 4.5 high thinking
The biggest question is, can they even pay for half of the deals they have been making?
I can’t help but wonder if their product orchestrated this deal.
Does this violate anti trust law?
> On October 1st OpenAI signed two simultaneous deals with Samsung and SK Hynix for 40% of the worlds DRAM supply.
The market doesn't believe they can pay for the Oracle cloud deal. Why do these vendors believe OpenAI can pay for 40% of the world's DRAM?
They definitely can, as anybody could, for a short enough period of time. The only ones betting they'll be able to sustain it for a long period are the ones paying the currently very inflated price for ram.
Altman was already unpopular. After this will he be able to show his face in Silicon Valley?
And all these data centers they want to build around the country. When consumers can’t get devices they want maybe they’ll fight even harder against these data centers being built in their back yard. He’s not making any fans with this move that’s for sure
I don't think Ellison is popular, but he seems to be doing fine with all his billions.
Imagine the outrage if OpenAI built their own fab or memory factory. Like back when Henry Ford built his own steel foundry.
almost feel like OpenAI's recent "fall" is a decoy setup by them intentionally.. something's cooking.. maybe they wanted to buy back their own shares at a lower price?
This will make AI even more palatable for the general population /s
Now this... was a really good move.
OP is also marginally underestimating the impact this move would have on Google's competitiveness - they are making huge gains prototyping at light speed; this will halt their AI hardware acceleration plans pushing them back into slower software development on ever aging hardware.
It also shows why Nvidia is not afraid of competitors coming out with new desgings that obsolete their hardware: what good are superior designs with no fabs to produce them?
every one of these things that make the deal "good" for OpenAI is a direct result of negative externalities for everyone else: competitors, consumers, and people who wouldn't care otherwise.
The article even says that they don't have an obvious plan for how to use the wafers they bought, and very clearly suggests that this is purely an anticompetitive tactic to force everyone else to eat a price increase that OpenAI doesn't need to face. It's clever though because if any regulatory agency starts asking questions (not that they would do that in the current USA political climate) then OpenAI can just say it's a strategic reserve, we have plans to do something with it, etc. etc. What are you going to do? Take them to court and force them to auction off some % of the stock? Set an industry-wide limit on wafer inventory? Fine them? You'd need to find some evidence that it was done maliciously, and good luck with that.
There are some negative elements of captialism that we might simply have no reasonable regulatory apparatus to deal with. Preventing indivduals and companies from having so much market power in the first place seems to be the only thing that can work consistently.
More anti sam anti AI propaganda, nothing dirty about this deal
As an AI researcher, I thought it was relatively well established (at least among my colleagues) that being pro-AI actually meant you were anti-Sam as well. He's the worst actor in the industry and has done an incredible amount of damage to its brand.