What’s stopping an American RAM plant? I would think we would have already broke ground.
Micron says Hi.
https://en.wikipedia.org/wiki/Micron_Technology
Micron Technology, Inc. is an American multinational semiconductor company that manufactures computer memory and computer data storage products, including dynamic random-access memory (DRAM), flash memory, High Bandwidth Memory (HBM), and solid-state drives (SSDs). Founded in 1978 in Boise, Idaho, Micron is the only major American computer memory manufacturer.[3] It is one of the “Big Three” computer memory manufacturers, along with the South Korean companies Samsung Electronics and SK Hynix.[4]
They have fabs in the US (and are building more).
https://www.micron.com/us-expansion/
Micron is accelerating its U.S. fab and technology investments, increasing its planned spend to more than $250 billion through 2035, driven by surging demand for memory in the AI era. These investments reflect Micron’s confidence in its technology leadership and are designed to support its long-term goal of producing 40% of its DRAM in the U.S. while creating additional good-paying jobs.
There’s been a lot of efforts to expand American fabs. Intel, Micron, and TSMC are all working on it (and I’m pretty sure most of those projects have at least broken ground at this point). But even with infinity dollars to spend, it takes a very long time to build these plants. The fabs are similar in size to a “normal” factory, but the entire interior is a series of massive clean rooms that are hundreds if not thousands of times cleaner than an operating room, filled with equipment that is simultaneously very large and very sensitive.
I remember reading somewhere that one of Intel’s newest process nodes involves drilling holes measured in hundreds of picometers into the part at some point. A hydrogen atom is roughly 62 picometers across.
Yes, all computer parts should be made by one company on an internationally disputed island. What could go wrong?
They also have a factory in Arizona.
Only for obsolete fab technologies for underpowered IOT chips. The fabs that make 2nm current generation chips are all in Taiwan.
It’s an interesting strategy if you’re Taiwan.
https://en.wikipedia.org/wiki/Silicon_shield
Silicon shield is the principle in geopolitical relations regarding Chinese unification that the semiconductor industry in Taiwan is important enough to global economic prosperity that it would deter an attempt by China to seize Taiwan by force, and if not would induce foreign powers, particularly the United States, to defend Taiwan in order to preserve the stability of the semiconductor supply chain.
Remember: you will not get anything from that ram. 100% will go to nvidia and amd for their AI. And it will block production capacity for other stuff.
Yup, this is just AI capture reaching the top of the supply chain.
“…to capture more AI spending”
It takes an already-existing memory manufacturer 4–5 years to do a new factory.
I don’t know how much longer it’d take a non-memory manufacturer, but presumably more.
I’m not saying that this would be bad, but I’d expect that it’d be some way in the future if it happens.
If anyone can pull this off in a shorter timeframe it’s TSMC. But that’s very ambitious even for them, of course.
and the reason manufacturers aren’t building many more factories is because they’re extremely expensive and there’s no guarantee the same demand will still exist in 5 or 10 years. What if the AI bubble bursts, the market is flooded with second-hand supply, and all the new factories aren’t needed?
Micron is building new factories in New York, and they expect the first factory there to be online by 2030 at a cost of $50 billion (which is more than their entire 2025 revenue).
I’m spitballing here, but one way it might be possible for TMSC (or whoever puts up the capital, which doesn’t have to necessarily be TMSC) to mitigate risk, given that this is Nvidia allegedly talking to them, is to build some kind of new memory specialized for AI use that would displace existing memory for AI use. Like, for neural nets, you don’t have the same constraints that you do on conventional DRAM. If you can have someone who builds neural net systems and a memory maker work together, you might be able to produce a new sort of memory that would be a better buy for neural net users.
A few things I can list off-the-cuff:
-
Access patterns on neural nets are predictable long in advance. That means latency doesn’t matter much (so having far higher latency than current memory may be fine) and requesting large blocks of memory might be valuable. Accesses will be sequentially requesting memory over a large range, so it also may mean that one could change how cooling works in the memory, if one has some guarantees on a single spot not being frequently re-requested a lot. A given byte of memory won’t be re-written, once written, for a long time, so you could relax how rapidly the hardware needs to be able to sequentially do writes to a given byte.
-
You have a lower ratio of writes to reads. I read one article talking about how HBM isn’t yet optimized for AI use on write/read bandwidth availability grounds. That’s talking about the memory interface, not the memory chip itself; they were just arguing for changing the ratio of bandwidth reserved for writes and reads. However, I would bet that you could also optimize the memory chip around this.
So you’d basically come up with, I dunno, “neural net memory (NNM)” or something like that. It’s not that you have to use it for neural nets, but that its characteristics would be optimized for neural net use. That only makes sense if there’s enough scale of demand for a single application to put engineering time in, expend the fixed costs, but there is presently a lot of demand, and might be more down the road.
That doesn’t make risk of demand dropping go away, but…it’d tend to shift that risk to conventional memory manufacturers, since if demand doesn’t meet expectations and you’re doing neural nets, a buyer would probably buy the specialized memory chip before the non-specialized memory chip. Still risk, but it’d be shifted to Micron/SK Hynix/Samsung.
At the start of the AI boom, we had hardware that was not optimized for AI use. It was generic random access memory with most of the hardware stack designed for rendering 3D graphics. I’d guess that it’s very likely that that hardware stack is substantially different from one optimized for neural net use, and I’d bet that Nvidia would be the most-likely party to be at the heart of attempts to re-engineer the hardware stack to optimize it, as that’s somewhere that they’re uniquely-well-positioned to make changes.
EDIT: I remember a while back when John Carmack was tossing the idea out there of using a loop of fiber-optic cable as a memory for AI use. That may not be the optimal design, but one does see that there is a lot of room to think about specializing the memory to the very different demands that neural nets place on memory rather than the demands that conventional RAM users have.
Not an expert on AI (or even software development), but geniune question: wouldn’t there still be the problem of silicon allocation involved?
Silicon allocation?
-
The kind of memory the current manufacturers are focusing on is specialized for data-center-grade GPUs. These GPUs are insanely power-hungry and require data-center-grade cooling. When the bubble pops and data centers go under, you’re not going to see consumer-grade RAM flood the market. So if TSMC wants to enter the market conservatively with the assumption the AI surge will not last, going into the consumer RAM side would be a better bet.
There’s a lot of regular RAM in the servers too though, since the servers can’t operate entirely from VRAM.
deleted by creator
We had this issue during the crypto boom and the manufacterers that added capacity then are glad they did.
What else could use up that RAM? How about standalone VR? We could get mainstream adoption by 2030.
The point is building factories sometimes pays off
I agree that more supply would be great and building factories can pay off. It’s still very expensive though, and some banks are growing skeptical of the ballooning costs of AI-related projects.
That’s why I don’t think we should limit the AI boom, give the supply chains a chance to pay for more capacity for the future
There are a few other reasons to stop building data centers.
Oddly it does say “adding memory back into its portfolio” which implies that they used to do memory, so they might have mothballed facilities.
It’s still a mammoth undertaking though.
Also, if they’re already producing on-die DRAM for SoC’s then perhaps that’s their entry point.
Not much more. They have the competence and colatteral
oh great, now everything else inside a computer will get expensive, too.
FCK NVDA
I’m waiting for the advent of 3d printable RAM. It doesn’t need to be good. I just need 8MB for my microcontrollers, that’s all.
That’s been done already with resin printed chambers, vacuum, and silicone seal at the top, which also serves as an indicator of the contents of the cell.
WAT. You’re telling me we can 3d print vacuum tubes?
Please do, this could make war less likely
