Hacker News Reader

Comment by rwmj | Hacker News Reader

Link avoiding cloudflare: https://archive.ph/oU8F1

The article is a rather poor one though. There have been several T-HEAD (Alibaba's chip subsidiary) designs before this one. The key issue for supply-chain security is where they are manufactured, which is still TSMC unless something radical has changed in China.

Of course China might invade Taiwan or might develop their chip manufacturing capabilities (only a matter of time IMHO). That would be real news.

None of this has much to do with the RISC-V ISA which is just a common standard. The article would be equally fact-free FUD if this was using China's own LoongArch.

View on HN

> The article would be equally fact-free non-news ...

It may not be surprising but it's not fact-free or non-news. The pace at which Alibaba is developing its RISC-V designs is significant. Removing one dependency on western supply chains is material development even if the dependency on TSMC remains.

There are tons of Chinese designs, and have been for at least 10 years. T-HEAD ones are most notable for how slow and buggy they are. StarFive (formerly SiFive China) and Sophgo (recently sanctioned by the US, using a variety of chip designs) are much more interesting. [Edit: & XiangShan]

Is TSMC reason enough to invade an entire country ? (I mean, sure the west don't really an excuse do to the same, but I wonder)

Can't they just "exfiltrate" or "compromise" or "steal" company secrets and establish their own base in mainland China ?

Genuinely curious as I know nothing of industrial espionage.

They've been doing that too (as well as attempting to steal secrets from ASML), but I guess having designs for building a fab isn't really the same as having the know-how from people who have worked on this for decades. The real danger from China invading Taiwan from a chip point of view is that they'd "sanction" the west and prevent us from getting any modern chips at all. Europe and even the US doesn't have leading edge fabs. Trump just disapplied the CHIPS Act too. The only other ones are in South Korea, so you can see where this is going ...

There is zero chance that anyone can invade a country and take its semiconductor infrastructure even vaguely intact if the defenders are disinclined to allow it.

A cup of coffee dumped in the machinery would destroy most of it beyond all hope of repair, let alone something fragmentary and full of nitrogen bonds.

The only sanction you could perform would be "no chips for anyone, anywhere".

And in such a, uh, kinetic, situation, I'm absolutely sure neither side will allow the other to retain working facilities that don't undergo very unfortunate accidents. Even a sudden power cut can be devastating. A foil balloon was enough to take out multiple fabs in Germany, and Samsung dropped 3.5% of global flash supply on the floor from 30 minutes down.

MAD doesn't come only in nuclear flavor.

For sure, but I guess China would hope they could capture the population, who would go on to rebuild the fabs with their know-how.

I wouldn't want to bet that you could do a $10 billion plus project to build the most complex facility in the world, in a tearing hurry and without complete documentation or vendor support, in a war economy and under international sanctions using a bunch of extremely pissed-off people and have the project go very well at all.

And I'd also bet that a translation of Simple Sabotage would be going viral within the week.

> A foil balloon was enough to take out multiple fabs in Germany

I didn't know about that incident; more details here:

https://www.datacenterdynamics.com/en/news/wayward-balloon-c...

TSMC is very difficult to copy. Optimistically it'll take you the best part of a a decade to build a leading edge fab. At which point you'll be a decade behind TSMC.

> At which point you'll be a decade behind TSMC.

Moore law is losing steam. TSMC itself is probably just 10-20 years behind physical limits. A decade from now TSMC will run out of room to be much ahead anyway.

Possible in principle, very hard in practice, even with full access to necessary asml machines and with a lot of experience and expertise without the difficulty of stealing incomplete information.

They certainly can’t “just” do that, no :-).

Source: see Intel, who have the money, experience, expertise, time, incentive and asml machines to do this and are making a big bet on gaining TSMC parity (at least). If the stock price is any indication, they are unlikely to succeed.

Disclaimer: I work at Intel but in an unrelated area. My information in this area is the same as an interested outsider. I do fiercely want them to succeed in becoming the only tsmc competitor.

>Is TSMC reason enough to invade an entire country ?

This make it sounds like they will invade just because of TSMC, but it is more like using TSMC as an excuse if it really happen.

China wants Taiwan, TSMC or not.

Reply on news.ycombinator.com