How would we do that? There's no historic precedent for that. Its fundamentally an information theory thing: what's the max amount of intelligence you can get out of 1 KB/MB/GB? There has to be a limit and I'm not convinced that it's far off.

And by the time the models that require 1TB memory will be better.

It's like people saying mobile chips are going to be better than PC chips... until they realize PC can be made of mobile chips too if that comes true.

The fundamental issue is that for these systems bigger is essentially always better (if affordable). So if we can squash something like Kimi K3 down to run on a 'normal' system, that just incentivizes devs to increase the model size until once again we're at the limit of what can be run.

I think the chipmakers are pretty securely set up here, the usual cyclical issues aside. Between the stubborn resistance of the data to being compressed without significant loss, and to a lesser extent the ability to do more with more memory (we’re surely at diminishing returns for LLMs already, but diminishing doesn’t mean zero) it seems like a fairly safe guess (I am not an expert on anything) that for the next decade at least you’ll still want at least 128-256GiB of video RAM to have a comfortably “SOTA-enough” experience. Then the other side of the rectangle comes from putting that VRAM in every consumer and office-worker device. Selling everything at a big premium to a few efficiency-obsessed SaaS providers is what you do when you’re temporarily supply-constrained. Sidelining the SaaS middlemen and selling to gloriously inefficient consumers who will leave their laptops closed for most of the day is the dream. This is probably not unrelated to the fact that the country which is currently making a big strategic push into memory manufacturing is also championing open models.

What do you base this prediction on? It seems very unlikely, unless by SOTA you mean the current SOTA.

And besides llm's memory requirements:

-HBM has a lot of room to grow(bandwidth/power/density). And an HBM memory takes 3x the amount of wafers of regular DRAM.

-HBM + better DRAM designs could increase random memory access speed by alot. I've seen research talking about 7x.

At what point it makes sense that half of consumer GPUs and most of the server GPU/CPUs start using HBM?

SoTA AI will move well past 1TB+ memory requirements in 1 year or so.

Also, you will be able to play with much more competent models locally. They will still feel like children compared to the adults living in the SoTA region.

Are you thinking of fundamental architectural changes (like the Transformer)? Or incremental (like MoE or GQA)? Are there specific neolabs or techs you are following that lead you to this prediction?

Indeed it seems quite possible we are one architecture breakthrough away from existing chip stock driving us all the way to ASI.

Nobody has done AFAIK the information theory to prove it’s not possible.

I'm looking forward to being able to buy CXMT chips for running SOTA models locally. :)

Yeah, you can directly print the SOTA AI model on the chip. I also believe this can be done on older process nodes. The most important factor would be speed. How fast can you go from new model to new chip?

Even if we do, we are going to need a lot of RAM for the billions of agents running everywhere.

You can get _better_ that SOTA AI performance by burning the weights directly to silicon, the problem is you're locked into that model

"we" being China?

NOPE. The opposite will happen. CXMT will flood the market thus making 1 TB models affordable.

that would be perfect timing

Hyperscalers are screwed, data center mania couldn’t even be completed during this massive spending spree, all while the people seemingly sitting on the sidelines are working on getting these things baked into the OS and chipsets in a way consumers wouldn’t notice

If the chinese government feels its critical to china's success then it will be backed like alibaba and wechat and others were

Tesla is kind the same. Everybody bought it, because it went up so much, less because of the product. I guess every AI would do the same.

South Korea has opened its financial doors to the outside world, and since SK Hynix is a publicly traded company, that's allowed capital to flow in and take control of the company.

> the investment mania in SK and it's really scary.

i mean, it shouldn't be scary, because their failure is their own, and i doubt there would be collateral damage outside of those high risk investors.

It's scary for _them_, but it's an interesting case study for an outsider.

The reverse might be worse for the US in the long run.

Right now, the US leads China in compute-heavy R&D because the US has access to more compute, and that compute has two components that China has limited ability to produce domestically: the GPUs themselves and the DRAM. A decent fraction of the balance of the plant is already produced in China.

If the US refuses to buy Chinese tech, then we may find ourselves paying more for compute than Chinese labs (DRAM first, but Huawei really really wants to catch up with the combined forces of TSMC, Nvidia and AMD) and falling behind on access to compute at acceptable prices.

If I were setting US policy, I would be trying to get the US-friendly companies to increase production and trying to encourage domestic production so that the US-friendly world can stay ahead. Getting held back by a DRAM cartel that wants to juice profits at the expense of output is not good for the US.

Trump owns micron stock.

No way he's letting Chinese chips in.

It's driven by the supreme leader's grand plan, like everything else there.

Look at DDR5-6000, it clearly is rising again since June.

> 8% to 9% of the worldwide DRAM market share in 2025

how much more do they need to ship to satisfy you?

The price is similar, as they all rely on same upstream suppliers. The quality is kinda meh.

No comment on the second part, but the first part is wrong. EU doesn't have a moat in process technology anymore (since 2010s). They have a bunch of legacy fabs, like 350, 180 and 130 nm but Taiwan, US, and China (and others like Israel, or Singapore) have these too. EU is severely lacking behind chip design, production and packaging capabilities for industrial, automotive, and space kind of niche applications since the heavy divestment from semiconductors after 2008 financial crisis and 2015 Euro crisis.

On top of this, once European giants like NXP (formerly Philips semiconductors), Infineon (formerly Siemens semiconductors), and ST are actively reducing their footprint in Europe and moving to India or SE Asia.

In short: No, EU has no future in those markets as designer or producer of chips.

This is actually at the core of our (EU) problem. Instead of thinking, "here is this new technology, how can we build new markets/products out of it", our industry leaders always think, "here is this new technology, how can we apply it to our existing industries".

You can see it a lot in AI, right now. AI4industry is not where the gains are right now. We should be competing on training foundational models. Once the foundational models from US/China are strong enough, they will do AI4industry and we wouldn't have any edge on them anyway because their tech is more vertically integrated and they have better insight in how AI works under the hood.

EU: Perhaps we should also spend a few billion on attracting american glue manufacturers so they open a factory and a half here so we can catch up in this chip making business.

> US happens to be the place where the managers

And you know.. the actual design for plenty of these chips. By that definition Apple is also an Asian company..

> with EU tooling

That is only a (vital) part of the supply chain, there are plenty of other components.

AFAIK the high end chips are made in Taiwan, that's not the same. Also there was an export control, so ASML couldn't sell the latest equipment to China.

> To be fair, all high end chips are made in China with EU tooling.

Did you mean Taiwan? Because depending on who you ask that is, or is not china.

The Chinese have been cut off from EU tooling for a while now. And their attempts to build out their own lithography setups have been a failure. Their 7nm process is, by all measures, not delivering the yields that one would need to make it really successful.

EU Chip industry : ASML, 2nd to only nvidia, 0.5T mkt cap

Even the EU thought the “EU Chips Act” was a failure which is why they are trying to re-invent it after spending billions https://www.theguardian.com/business/2025/apr/28/eu-microchi...

I don’t think the goal there was to sit around praising past accomplishments. You always have to keep building.

And that’s about it..

In relative terms there is no innovation happening outside of that and some other smaller companies providing highly specialized components that are a part of the supply chain.

There isn’t just much if anything to brag about. ASML is great but it’s effectively a monopoly (for the time being) due to a large extent some correct historical choices.

ASML doesn't make chips.

Yeah the audience here has significantly shifted.

Based on the empirical evidence we have ASML succeeded in spite of the EU not because of it (so “they” is bot exactly fair). There is not much EU has going on besides that when it comes to growth and innovation.

It’s a bit like Novo Nordisk, they had a huge win and then just focused on making as much money while it lasted while more innovative companies ate up their market.

> And they mostly stole their technology from hiring Tesla factory workers to moonlight for them and share the design, technology, tooling, and RE the Tesla finished product.

Did they successfully steal Tesla's low-cost EV plans too? Using the traditional definition of "steal" (depriving the owner of their property.) The argument that all China is good for is stealing falls apart[1] the second Chinese products become superior (see battery chemistry).

Perhaps Tesla should pay Chinese EV designers and PMs to moonlight for them to shorten their model development cycles.

Domestic Chinese EV competition is fierce, if one can survive or thrive in that boiler, dealing with VW, Ford, or even Tesla is a cake-walk.

1. Nevermind the growing body of fundamental research papers published. An area the US is turning it's back on because some people are gassing themselves up on they idea they are intrinsically, and immutably great/superior, therefore there is no cost to wrecking stuff to "own the libs"

It's this. Wall Street won't allow american companies to make long term decisions and long term planning that cost money in the short term, because they want all that money immediately now. Like, right now now. If an executive promised they could deliver big profits in 5 years by spending the next two investing their profits, Wall Street would fire them and put in someone who will happily spend all the profit (plus debt) to do stock buybacks and dividend disbursals.

It's even worse, the entire ideology is that industrial policy can't, by definition, be better than what the market would decide without interference. So every kind of actual strategic decision making (i.e. investing a few hundred billions in the industries of importance for the next decade(s), or at least in the infrastructure supporting it) would actually make matters worse. That is essentially what neoliberalism is all about.

This is core reason why the German industry became so irrelevant in the last 2 decades: The rail infrastructure became pseudo privatized and now is in shambles (IIRC only about 60% of trains arrive on time). The PV industry was axed by the conservatives about 15 years ago and now is completely dominated by China. Even now, as the other EU countries demand strategic investments - and again the German conservatives say this ought to happen without any new debt. "We can only spend what we have", to quote Merz. Imagine China saying this about chip production, or even Sam Altman about compute hardware...

The EU isn't monolithic, you could probably do this fairly easily in someplace like Poland where there isn't a bureaucratic hell and a large supply of skilled workers and engineers. But of course, Poland and co. are the "lessers" in Europe where you outsource for cheaper labor, a prestigious thing like a cutting-edge chip fab would need to be built somewhere like Germany or the Netherlands, and so it never shall exist.