Comment by KK7NIL
3 months ago
The best China has is an internationally uncompetitive "7nm" fab and that's the best they'll have until they can manufacture EUV machines domestically.
So the EUV blockade has absolutely been effective and the fact that the PRC is paying so many shills to convince westerners otherwise just shows how behind they are.
I noticed that people love pointing how far AI field has advanced in a few years and extrapolate next few years. While at the same time being dismissive of Chinese semiconductor manufacturing process. In similar vein I also remember claims that TSMC Fab in Arizona can never work, and yet it does. So I don't know man, I wouldn't underestimate what a billion of enterprising people can do. Especially when paired with the system that has a pipeline of funneling smart people into elite schools.
Underestimating China seems like a really, really, really stupid thing to do.
I don't think the US is underestimating China... I do think that the US is preemptively shoring up a domestic posture against long term changes. It would be a pretty bad strategy to continue to outsource everything and continue to see a massive trade imbalance with the outside world for a prolonged period of time.
4 replies →
Yes, we are doing a bad job of updating our priors.
Is that sarcastic? Isn’t underestimating by definition a bad thing?
3 replies →
Perhaps the USA feels that it has a reputation to downhold?
[dead]
Okay? There's a lot of chips you can make that aren't the cutting edge. You don't need a 4090 to do AI, as evidenced by all the AI we did before the 4090. You definitely don't need a (random Intel chip) 14900HX to do general-purpose computing, as evidenced by all the general-purpose computing we did before the 14900HX.
For that matter, the 14900hx was already based on a refined 7nm production process, which China already has started using, though maybe not as effectively yet. As you mention, prior to the 4090's 3090 was on an 8nm node, already behind current China capabilities.
If each node provides a 10-15% improvement in power, performance and area, how many of those need to compound until your already uncompetitive 7 nm is 10x less efficient, slower and more expensive?
Being behind doesn't mean they're permanently stuck where they are today - but aren't our processes running into the wall of soon trying to make transistors smaller than an atom?
3 replies →
You are ignoring the possibility of technological disruption.
Apple disrupted Nokia and Blackberry. ARM is currently disrupting Intel.
What if someone lands on a break-through using a completely different tech: what if X-ray lithography [1] becomes viable enough that they don’t have to acquire state-of-art EUV machines from ASML?
[1] X-ray lithography was abandoned in the 80s but it is being revisited by Substrate https://substrate.com/our-purpose. They are an American company that hopes to make it commercially viable by being cheaper and far less complex than EUV.
Substrate is a scam; their marketing is misleading and they have yet to answer to the fundamental reason why X-ray and e-beam failed over 40 years ago (despite it being generally agreed they were the future of litho and optical would soon be dead): writing one line at a time is extremely slow compared to optical which can scan a whole reticle in a fraction of a second.
E-beam is still used for making DUV/EUV masks where the low write speed can be tolerated but no one in the industry thinks it will replace EUV in the silicon litho steps any time soon.
But lay people eat this crap up and journalists turn a blind eye either because they're literally paid PRC shills or because clicks are everything now a days.
I think you're general point is completely true, but Substrate is a bad example, since the people running it don't appear to be semiconductor experts and it's probably a fraud.
Apart from gaming and llms, most of the chip applications including all of military and consumer electronics is more than happy with 7nm process, whatever that means (proper nanometers those ain't).
I know some people live in the IT bubble and measure whole reality by it, but that's not so much true for the world out there. They have ie roughly F-35 equivalent, minus some secret sauces (which may not be so secret at the end since it seems they stole all of it).
You are making a mistake of thinking of them as yet another russia, utterly corrupt, dysfunctional at every level and living off some 'glorious past', when reality is exactly the opposite.
So, you're saying that China has chip fabrication capabilities which are on par with the world cutting edge as of 2018:
https://en.wikipedia.org/wiki/7_nm_process
not too shabby of a fall-back.
No, they don't.
Their "7 nm" relied on multi patterning DUV which leads to restrictive design rules, more steps and masks and lower yields, which is why I put it in quotes and said it's uncompetitive.
The last DUV node was 10 nm, that's the best logic node they have which is comparable to TSMC/Samsung/Intel's 10 nm.
> that's the best they'll have until they can manufacture EUV machines domestically.
And how far out is that?
> And how far out is that?
These guys have a 100% market share https://www.asml.com/en/products/euv-lithography-systems at the 'extreme' end and, obviously, everyone else is trying but haven't really shown much promise.
Here's a good background article on the topic: https://www.economist.com/science-and-technology/2025/03/12/...
> everyone else is trying but haven't really shown much promise
What was the incentive/funding for their attempts? In a non-national-security scenario it makes sense not to try too hard because you can just buy ASML's solution.
With China it's a bit different, if they decide it's a matter of national security and pour Manhattan-project-levels of money/resources into it, they could make faster progress.
16 replies →
If you ask PRC shills, it's just around the corner because this one Chinese lab demonstrated a very small part of the system. And a surprising number of westerners fall for that crap.
My guess is that it's at least 10 years away, but that could obviously change depending on what resources they're willing to commit. But even at that point they'll be 2 decades behind ASML's EUV tech so it probably won't be competitive.
> If you ask PRC shills
GP must have been asking for the non-PRC shill opinion.
> My guess is that it's at least 10 years away,
That doesn't sound at all like a lot. China has a uniquely effective industrial espionage... industry, combined with a very thick geopolitical skin and disregard for international demands. This helps accelerate any process that others have already perfected.
We'll start to see the real deal if/when China eventually catches up to the leaders in every field and the only way to pull ahead is to be entirely self propelled (you can't take advantage of someone else's draft when you're in front of the pack).
I think you may underestimate the ability of China to abuse industrial espionage at scale.
5 replies →
So far only one company in the world has successfully accomplished it, so the answer could be "a very very long time".
According to this video (Asionometry - guy from Taiwan, hardly a PRC shill) Chinese EUV are now tested in Huawei factories and should come into production in 2026.
https://www.youtube.com/watch?v=rIR3wfZ-EV0
“Huawei has 208,000 employees and operates in over 170 countries and regions, serving more than three billion people around the world.”
https://www.huawei.com/en/media-center/company-facts
“The company's commitment to innovation is highlighted by its substantial investment of 179.7 billion yuan ($24.77 billion) in research and development (R&D), accounting for 20.8 percent of its annual revenue. Its total R&D investment over the past decade has reached 1.249 trillion yuan ($172.21 billion).”
https://news.cgtn.com/news/2025-03-31/Huawei-reports-solid-2...
They have the incentive, the government backing, exist in a mature ecosystem of tech rivalled only by the US, … If any corp can do it, Huawei can
I rewatched the whole video and did not find where he said that. Quite the opposite, he says Chinese EUV academic research is at 2005 levels and is rather unimpressive.
> the fact that the PRC is paying so many shills to convince westerners otherwise just shows how behind they are.
And yet, it's anti-PRC shills that are all over social media. Go figure.
They can just throw power at it, you're delusional if you think it's going to hamper them even mid term.
My understanding, which is not complete, is China has done some amazing things optimizing training on slower chips.
Which is cool, but there are limits to the number of times you can do that.
At the end of the day, the little man has to flip the switch.