Nvidia and its partners built a system to bypass U.S. export restrictions

That's a bit of an unserious argument, to put it mildly. When people say "China" here, we're talking about a specific subset of bad/adversarial actors, whether in government or (I guess in the case) the drug trade. I have no problem with the vast majority of that 1.4 billion.

Tinygrad is doing this too.

I think you misunderstand, you don't need 90% of workers to act against a factory to sabotage or shut it down. You need one to do significant damage. Not percent, one employee.

If you think about where you work, I'm sure you can quickly figure out how you could cause significant harm to it. I'm sure if you think for a bit you could figure out how to do more and maybe how to avoid getting caught or at least caught quickly.

> Reality is, as with all war, collaborators will be richly rewarded, sabeteurs will be shot.

Not correct at all. Collaborators do not typically become richly rewarded. There are examples, but there is not a single instance of an occupied country where that country's people all got rich corroborating. Nor even high rates of collaboration.

Thing is, people aren't very different from you or I. If another country performed a hostile takeover of your country, do you think you'd happily collaborate? Maybe you would. Would you do so without reservation? Do you think your decision will be common? Have you considered that this is a hostile takeover and there's a very good chance that at least one of your friends or family members has been killed.

There's an old saying that was popular during the initial invasion of Iraq:

  How do you create a terrorist?
  You kill his brother.

In case you missed the context, it is a critique on the invasion itself and how the actions being taken were creating more adversaries.

Yes, people caught will be shot. But this creates two types of people, not one. Those that are afraid and reluctantly comply and people who are afraid and are catalyzed to fight back. Don't believe me?

Look at the history of literally any occupation effort. Hanging onto occupied territories is very difficult. Take Ukraine as an example if you want to understand things from a real life evolving situation. There is still resistance in Donetsk, resistance in Luhansk. Hell, resistance has not stopped in Crimea, a decade after occupation.

If you think the Taiwanese will just roll over and comply, then I think you are being naive. I think you haven't even though of what the impact of losing a loved one will do to you.

  > PRC is losing a finger, west is losing the hands and feet.

Again, these are not turnkey operations. I'd also suggest you attempt reverse engineering something before making such strong claims. I'll even make the bar low: just reverse engineer hardware. If you can do that, you're 20% of the way to success! (even adding the software won't get you to 90%). China is losing a lot more than a few fingers, and taking Taiwan doesn't get them Samsung. The Koreans aren't big allies of China (btw, they still hold a grudge with Japan more than 70 years later). It doesn't get them Intel. It doesn't get them Broadcom, Qualcomm, Hynix, ASML, or even AMAT. HiS isn't even close to Samsung. Yes, global chip production will fall and be hit hard. But remember which country buys 54% of chips...

  > Unless there's replacement/substitute, which so far PRC is the only country trying to replicate entire supply chain

https://en.wikipedia.org/wiki/TSMC#Arizona

> this is forgetting that the US exists.

I did not forget this. I think you overestimate what the US can and is willing to do there, and also what the incentives would be for people to escape to the US. This is not Vietnam in the 1970s, staying right there may very well be the preferred choice, even if China takes over. Especially if they make a very nice offer in advance. Should Trump win that may look even worse.

If PRC gets TW fabs intact (which presumes invasion successful), PRC semi engineers will be overseering to filter for loyalty. This isn't unique historic problem, unfucking captured labour production from wartime spoils is a process. Nor do you need 100% TWnese loyalty, only group of compradors willing to be rewraded to drive internecine fracturing, of which there always are.

>occupation

Borders change through history because occupation frequently works. Chinese history is literally continuous record of various groups being occupied and absorbed. Missing in this context is PRC:TW has 50:1 manpower difference vs 4:1 in RU:UKR. Note that Tibet and Xinjiang has been thoroughly securitized, XJ have population approximate of TW. PRC simply has a lot of manpower advantage... millions of PLA/PAP to spare for occupation... and TW isn't a frontier province with difficult land logistics (PRC couldn't tame XJ/Tibet until they built out expensive rail infra) than ferrying troops/supplies via water (again if invasion successful). No one expects TW to rollover, but taming 25m people is well within PRC abilities.

>operations

PRC is basically the only actor with sufficient talent generation able to, and with intent to build out ENTIRE indigenous semi supply chain. Reverse engineering hardware is only piece of puzzle. Yes semi is particularly difficult, but PRC indy policy has fairly proven record of being able to indigenize tech within reasonable time frames, some take longer (i.e. turboject, semi). Large % of PRC of semi imports goes towards export, of which US/west captures significant share (IP etc)... PRC is going to lose $10 in iphone assembly fees while west loses $100s of BOM in semi components... the actual accounting is where fingers vs limbs becomes obvious. If Samsung can easily replace TSMC they wouldn't be shutting down leading edge semi lines right now (50% by end of year), not to mention there's chance Samsung fabs would go boom in broader TW conflict.

>TSMC Arizona

TSMC Arizona likely will still depend on many sole source suppliers on TW short/medium term (5-10 years). Talk about TSMC TW being unsustainable due to foreign imports of hardware/maintence etc also rings true for fabs on CONUS or elsewhere, good chance they'll stop operations without TW exclusive inputs that I'm sure many are trying to substitute as we speak. But again PRC is likely only country with industrial base to replicate entire semi supply chain in short/medium time frame... PRC only actor without projected semi talent shortfall. Hence depending on timeline/rate of indigenization, west will likely lose bulk of leading edge node advantage that where western incumbants derrive disproportionate value capture and significant net losers relative to PRC.

Even if TSMC Arizona keeps chugging along, entire US Chips Act isn't projected to capture more than 10% of leading edge by 2030. US high tech losing 90% of leading edge hurts much more than PRC who are already largely denied ability to capture leading edge shares. Meanwhile PRC is rapidly expanding mature nodes, so we're looking at potential scenario where PRC continues to hobble along on 14nm+ while west loses 90% of leading edge nodes and ~60% of mature nodes (by 2030 PRC projected to have ~40% of mature nodes, TW ~40% of 60% remaining, i.e. 2/3 of mature nodes in western bloc). This dramatically closes semi production gap in PRC favour, west high end node gap becomes marginal, while PRC potentially 2:1 or 4:1 (if TW mature nodes captured) lead in mature node production. That's the numbers that determine winners/losers, or in this case relative loser (PRC fingers) / big losers (western limbs). PRC capturing no TSMC fabs but denying west said fabs is already nightmare scenario for west, PRC capturing is just bonus leverage, i.e. offer west continuity in global semi supply so everyone can transition into bloc supply chains for conceding on TW.