Comment by DivingForGold
10 hours ago
"AI has a physical dependency in Taiwan that can be easily destroyed by Chinese missiles, even without an invasion" ?
Arguably false. Why do you think the US has encouraged TSMC foundries, now inside Arizona ? It's obviously to protect against the scenario that China takes Taiwan. In that case, give it 6 months or less for US TSMC foundries to produce the finest. China taking Taiwan will likely not result in the CCP getting any technology, certainly Taiwanese have "contingency plans" to vaporize all tech in the event they are invaded.
That's over-optimistic.
> Why do you think the US has encouraged TSMC foundries, now inside Arizona ? It's obviously to protect against the scenario that China takes Taiwan. In that case, give it 6 months or less for US TSMC foundries to produce the finest.
American business culture works pretty strongly against "give it 6 months or less for US TSMC foundries to produce the finest."
I think it's most likely those fabs will stagnate and the American MBAs running them will just milk them for short-term profits. Why invest in technology when you can buy back shares? After all, your only goal is number-go-up.
> China taking Taiwan will likely not result in the CCP getting any technology, certainly Taiwanese have "contingency plans" to vaporize all tech in the event they are invaded.
Not necessarily. Technology isn't so much the machines, it's the know-how. TMSC employees will still need jobs, post invasion, and I'm sure China will pay them very well. Some fraction will go to work for Chinese fabs, and teach them TMSC's tricks and knowledge.
>American business culture works pretty strongly against "give it 6 months or less for US TSMC foundries to produce the finest."
>I think it's most likely those fabs will stagnate and the American MBAs running them will just milk them for short-term profits. Why invest in technology when you can buy back shares? After all, your only goal is number-go-up.
The same American business culture that produced Apple/Amazon/Alphabet/Microsoft/Meta/Tesla/etc? They seem to invest quite a bit in "technology".
> The same American business culture that produced Apple/Amazon/Alphabet/Microsoft/Meta/Tesla/etc? They seem to invest quite a bit in "technology".
Yeah. There can be sparks of innovation, but the overall trend seems to be squandering advantages for short term financial gains. And honestly, some of your examples aren't great: Telsa's probably going to lose to the likes of BYD (but for unusual leadership reasons); Microsoft seems to be losing former capability, which is reflected in many of its products (https://www.windowscentral.com/microsoft/windows-11/windows-...); and I don't think Meta has done much besides sell ads.
IIRC, Chinese companies, on the other hand, tend not actually make very much money for their investors.
1 reply →
No, the same idiot MBAs that almost destroyed Intel and Boeing and General Electric.
2 replies →
TSMC's Arizona fabs (edit: need to qualify this with "today" and in the near future) are wholly inadequate to shift wafer volume out of Tawian if that were to ever happen. TSMC themselves have been candid about this - both the fact there is insufficient skilled labor and insufficient economics (materials supply chain, construction costs/process, subsidies, OPEX). If TSMC was serious about this they'd have invested heavily both in staff pipeline (university programs and hiring onramps), domestic executive function and supply chain - aside from taking subsidies and building tiny fabs that trail their Taiwan process nodes considerably, they've done little to diversify their fabs.
90% of TSMC's capacity is still in Taiwan. A substantial amount of global high end chip capacity is also in South Korea and Japan, which would likely get pulled in.
A war would not wipe out chip production, but the squeeze would be immense for many years.
> A war would not wipe out chip production
It probably wouldn't but it definitely could.
>China taking Taiwan will likely not result in the CCP getting any technology, certainly Taiwanese have "contingency plans" to vaporize all tech in the event they are invaded.
He didn't suggest anything like that, did he?
I don't even know what it means. "even without an invasion"? The author think China will destroy TSMC just because? To slow down AI progress?
> if we got to a situation where only the U.S. had the sort of AI that would give us an unassailable advantage militarily, then the optimal strategy for China would change to taking TSMC off of the board.
Lmao it's not. The author doesn't know what they're talking about at all. Let's be realistic: the current TSMC technology will be accessible to China, likely via espionage. The question is just how soon. It has already happened before. China's 7nm process was developed with the help from one of the highest level ex-TSMC researcher[0].
[0]: https://en.wikipedia.org/wiki/Liang_Mong_Song
If cutting edge were a hard requirement then given the lead times involved I think the author would be correct. However I think there's a fundamental error in failing to account for the fact that you don't need cutting edge chips to do AI. Sure it makes it cheaper and faster but it's absolutely not a requirement. You could train a state of the art model on cluster of 12+ year old boxes (ie Intel's 22 nm and DDR3) but if you want to get the job done in a similar timeframe you're going to pay out the ass for electricity. Your research pipeline would necessarily be narrower due to physical and monetary limitations but that's not the end of the world.
That’s like saying you could train a state of the art model by hand, and it’ll only cost you a lot of man-hours.
Realistically, to train a frontier model you’d need quite a lot of compute. GPT4, which is old news, was supposedly trained on 25,000 A100s.
There’s just no reasonable way of catching modern hardware with old hardware+time/electricity.
1 reply →
I suspect the bottleneck on 12+ year old hardware wouldn't be power but the interconnects. SOTA training is bound by gradient synchronization latency. Without NVLink you hit a hard wall where the compute spends most of its time waiting on PCIe or ethernet.
1 reply →
Mainland China has no interest in destroying the fabs in Taiwan, quite the opposite.
Taiwan might go scorched Earth and destroy them but that sounds more like a threat to foster US' support.
For the US the threat is either destruction of the fabs or China leverage against them if they get to control the fabs.
Hiring people isn't espionage. Key talent leaving a dominant manufacturer for a paycheck at a struggling competitor and bringing their knowledge is just about the basis of capitalism.
In 6 months TMSC US foundries would grind to a halt without consumables from TW who still has many sole source suppliers. The reason for CONUS TSMC fabs is because that's all US industrial policy can really manage, CHIPS never pretended to be able to reshore the entire semi supply chain (unlike PRC) and until US does, or at least reshores everything sole source off TW, will remain exquisitely vulnerable.
Also TW politicians have objected to notion they'd vaporize their fabs / golden goose. That meme started by US Army War College + Colby who said US should blow up the fabs, as in it shouldn't be a TW decision. Which TW have rebuked said they will defend against US attacks. Also other shenanigans like when US suggest they would paperclip TW semi engineers, and TW basically said there's no way they'd send semi engineers to safety before children. AKA TW not retarded, they know not to toast their golden goose, because golden goose for PRC still gives TW leverage even if TW forced to capitulate. They'd still rather be wealthy semi producers than pine apple farmers under PRC.
The Arizona fab's not supposed, per wiki, to start producing 3nm until 2028. Are you suggesting that the only reason it can't be done by summer is a lack of motivation and resources?
>In that case, give it 6 months or less for US TSMC foundries to produce the finest.
It's really a blind belief in american exceptionalism that makes you think this is even possible.
No, the chip factory that has had dozens of years of experience and local talent scaling up to make the most complicated products in human existence doesn't magically get up to par in 6 months. At best in 6 months they've figured out how to be less sensitive to vibrations and reach a low yield. The US doesn't have the trained workforce for this job, nor the infrastructures _around_ the fab (specialized hardware, electronics and engineering schools, various bits and bobs).
US TSMC doesn't get properly running in less than 5 years, and even that would be a miracle. You're also assuming that US TSMC has the current N2P or even N3E processes, and that agent orange doesn't burn bridges with europe hard enough that ASML stops selling to anyone related to the US.
The overwhelming bulk of production, in a massively oversubscribed industry, is in Taiwan. If Taiwan's production went offline, there would be enormous turmoil.
"certainly Taiwanese have "contingency plans" to vaporize all tech in the event they are invaded"
Given the way things are going, a rationalist would surmise that Taiwan is likely in talks with China for a peaceful reunification, Hong Kong style. The old way is very much over, the US is a worthless if not negative-value ally, and it's pretty clear to every living human with a functioning brain that this is going to be China's century.
Indeed, the article casually says "Taiwan is claimed by China, which has not and will not take reunification-by-force off of the table", which is technically true it isn't contextually informative. For those not fully up to date on the history of this conflict, for decades Taiwan claimed all of China (and most Western countries treated Taiwan as the singular government of all of China), and held out for reunification-by-force. This isn't as simple and straightforward like Trump's "we have a military ergo we get to steal better countries because they make us look bad".
https://www.youtube.com/watch?v=pF-ZN11DRSE