https://www.wsj.com/articles/china-wants-a-chip-machine-from...
the company is probably as important as TSMC in terms of geopolitics now
This might actually be a good thing for China. In a "necessity is the root of invention" kinda way. Ultimately this might be good to the world too, competition is generally a good thing.
China can say they have a priority on domestic semiconductor production, and they can throw obscene amounts of money at the problem.
But ultimately, something that takes 10+ years to develop isn't appetizing for China's current political climate. It needs to be the best, yesterday.
Consequently, they'll continue throwing money at whoever promises results next year, and starving whoever says it can't be done but offers a realistic plan to eventual dominance.
Capitalism sucks in many ways, but decentralized command economies (national government commands, local supplies) are about the worst of all possible worlds for high knowledge, long lead time, heavy research industries.
Which is why you saw the Soviets do pretty well at heavy industry, but not so well at computers.
Yet DARPA and company are behind like half of modern IT technology.
>Which is why you saw the Soviets do pretty well at heavy industry, but not so well at computers
Soviets were fairly behind the US in electronics to begin with, and they had massive production problems everywhere, including heavy industry(turns out that setting every price and production quota from Moscow wasn't really effective). It's just that heavy industry was easier.
Program managers, selected from industry and academia, given open remits for 3-5 years, strictly capped when the deadline is reached.
Of course, it requires the new external hires be given enough influence and autonomy. It remains to be seen if Chinese firms are able to integrate international talents effectively.
When the DoD was deeply involved with SV instead of just purchasing top binned COTS? At that time they basically were operating with hardware that far ahead, but that was always relative to the general consumer, it didn't stop companies from paying insane prices for marginal improvements that only HFTs and research labs would care about. Nobody would have bothered trying to bring something so expensive to the broader market - remember the multi-core games Intel played for years before AMD started knocking bread off their table?
Also, are you implying that the lack of leaked supercomputing documents is evidence of a commercial technological parity, or did I miss a big reveal (totally possible, haven't been paying attention)? Because as I recall he didn't pull a Manning style dump of operations-level intel (regardless of legality) - he instead coordinated with Greenwald to expose illegal strategic programs (and occasionally a debatable legal one that the DNI lied about in the prior news cycle - which was pretty funny). So I didn't expect any leaks about bleeding edge crypto hardware - because most domestic communication wouldn't need space age tech to break (A5/1 has always been hobbled).
For example, the surface they are printing the chip circuits on has layers of material, say, tungsten, that have thicknesses measured in atoms. As in, 'I need a layer of tungsten 100 atoms thick'. At that scale, the surface is 'bumpy' to the point that you must re-focus each time the wafer/lens moves to the next region on the wafer. This is a mechanical adjustment at atomic scale that has to happen very fast, or you'll never get enough wafers through the machine to pay for it. The solution to just this one subsystem of moving and refocusing required a decade of R&D and dozens of scientists and engineers to get fast enough, reliable enough, repeatable enough to be useful in a production setting.
Yes, others can make a system like this, with a similar amount of R&D spend and personnel and time. They would have to move faster than ASML to ever catch them. They'd have to budget 20 years, 100s of staff and a very large capital budget. In 20 years, ASML will be that much further down the road.
Here is a short animation of the optics:
Which works until China invades Taiwan. Then what?
I'm not an expert, but I was surprised to learn that GPS chips aren't as simple as I imagined. E.g. to get a "normal" (civilian I guess) licence you have to manufacture your chips so that they shut down if the object is moving too quickly (otherwise it could be used for missiles).
It can be done for probably less then a thousand dollars, even.
It's also pretty trivial to detect the simple falsification schemes, but outside of exotic military and research projects, none of the easily available GPS SoCs bother.
On that you are very right.
I'm often amazed by GPS, or any of the other GNSS technologies really.
A fleck as big as a thumbnail which can communicate with multiple satellites wizzing-by to lock your location within 10s of meters. The fleck contains the antenna, the RF frontend, signal locking and tracking circuit, and a small computer to provide the navigational solution from the observables. And you get all this wonder super cheap.
The whole system, the space vehicles, the ground segment is of course very expensive, but somehow the militaries of the world decided that they will you use all of that for free. If you would write this in a sci-fi I would trash your manuscript immediately. Except somehow it is not science fiction. It is reality.
GPS capability delivers incredible military advantage. (See: Gulf War)
Consequently, US government funds constellation. Because money is no object vs the military advantage.
Because of the nature of GPS, receivers (devices) can more or less be iterated independently of transmitters (satellites).
US government allows civilian use (with reduced accuracy) for economic and geopolitical reasons.
Commercial market rapidly iterates on the device technology (smaller, faster, cheaper).
Transmitters are upgraded and enhanced, as expanding commercial market now justifies expense. Eventually accuracy limitations dropped.
We end up at today.
It’s also worth considering that the cost of living in the US has been lowered because China has been the factory for our WalMart / Target / IKEA goods. If we didn’t open up China to the US / WTO markets, the world would be a very different place and that counterfactual is pretty difficult to calculate.
Also cost of living for who? The now-unemployed swaths of the country certainly didn't see their cost of living go down in terms of hours worked.
Yes, there are people who will lose jobs for all sorts of reasons, including competitive advantage of different regions (eg China and Mexico have labor that is far cheaper than Americans are willing to work for) or the euphemistic “creative destruction”.
I don’t feel sorry for horse buggy whip salespeople who had to search for a different purpose in life. They already won a lottery by being born in the USA in the most affluent time in history. The economy doesn’t owe anyone anything.
The only thing I wished we as a country had done during multinational trade pacts was not to put the onus on individual employees for proving jobs lost to offshoring; there was moral hazard and the evidence to prove causation was hidden from those who needed the evidence. Other than that, these trade pacts simply accelerated the changes that were going to happen anyway. It’s better not to live in a one-industry town where the winds of change can decimate the Un-diversified local economy.
Your usage of the term has a loaded definition. You're trying to thread a needle between the nominal cost of living (which has increased) and effective cost of living (which has increased for some), presumably by appealing to CPI. But you can't appeal to a measure that's from the exact paradigm I'm criticizing - of course things look consistent and reasonable from inside that paradigm (but its results on the real world are problematic).
> I don’t feel sorry for horse buggy whip salespeople who had to search for a different purpose in life. They already won a lottery by being born in the USA in the most affluent time in history
I'm not arguing that we should have held back offshoring, and you can spare me the race-to-the-bottom privilege indictment.
> The economy doesn’t owe anyone anything
This is a passive voice, while I'm criticizing a specific actor. "The economy" didn't decide that prices have to go up year over year, and in fact they would have naturally gone down with offshoring. Rather, the government decided that prices still needed to go up year over year, despite the gains made from offshoring.
If prices had been allowed to decrease, then the communities that lost manufacturing jobs would have at least seen some benefit from goods costing less as their own expenses would have gone down. But rather than the promised benefit of "cheaper goods" being realized throughout society, those gains were collected centrally far away from the people most impacted.
Savings from China trade are fairly trivial, calculated in a few hundreds dollars per family per year. And if the US need a cheap factory hub there's Mexico.
It would not be difficult to conduct a campus-wide evacuation then have the Taiwanese Air Force destroy the facility with targeted weapons. It might also be the case that the facility's precise coordinates are well-known to Taiwanese military's artillery units and the entire place could be reduced to rubble in minutes.
Because so much trade is done in dollars. And because companies globally want access to U.S. financial markets, money transfer systems, companies and markets, the U.S. has the ability to tell even foreign companies what to do.
That's why Europe was frustrated when Trump reinstated sanctions on Iran. And had to try to figure out ways of bypassing them. (https://www.google.com/amp/s/www.voanews.com/middle-east/voa...)
The U.S. has incredible, unexpected power over global companies and trade.
True, but the more it uses it for self-serving purposes the more the rest of the world tries to find alternatives, weakening that soft-power in the long-term.
The Taiwanese will destroy all the fabs before China can take them. We already have a chip shortage, can you imagine what it will be like to cut current supplies by another 40-50%? The resulting sanctions alone would cripple China.
Hence, status quo will remain, and Taiwan will continue to be a prop for nationalist speeches in China and the US defense industry will continue to make a mint off selling them old gear.
> imagine what it will be like to cut current supplies by another 40-50%
And despite everyone's aforementioned desire to preserve TSMC, ultimately this is still a trivial amount if you consider that the East Asian semi supply chain accounts for 75% of global IC share and is completely within PRC missile range if there was ever an expansive east Asian theatre over Taiwan. Losing TSMC in war over TW will set the world back 1-2 nodes and 10 years, losing regional fabs and supply chains (in SK or JP) in a broader war over TW will knock out decades worth of capex. Global semiconductor industry is a 500B market, trillions in terms of industries affected, most of which benefits the west. PRC has massively more leverage in semi mutually assured destruction, so much so that TSMC is ultimately expendable if PRC is ever forced to act upon TW, because losing TSMC doesn't even scratch the surface of semi disruption PRC can cause US/West, who again are disproportionate beneficiaries of current IC status quo.
How would the PRC be forced to act on Taiwan?
If China invades Taiwan before enough TSMC plants are built in America to meet American, and by extension - Western - demand for electronics, then it's war. Period, end of story.
There is no fucking way in any timeline in any alternate universe that Western electronics companies can allow China to put their businesses on hold, because it isn't just a bunch of AMD and Qualcomm CPUs... it's every automotive company in the Western world. It's every appliance manufacturer in the Western world PLUS South Korea.
China will go right back to 1820. Damn near dead fucking last on the world stage.
The Communist party leadership would probably kill Xi Jinping before allowing him to make such a monumental mistake. And nuclear war isn't an option either, because all of China will end up completely uninhabitable if even a single major American city gets targeted.
And it won't just be "America vs. China" either, it'll be "Every Western nation vs. China". And China will lose. Badly.
Car/appliance manufacturers do not use their cutting edge chips. The biggest use-case for these chips are new iPhones and iPads. Nothing critical.
If China takes control of chip supply the US will have to play along and try to apply soft power, never war, because China is a nuclear nation with 4 people to every 1 in the US.
It hasn't even been a month since Afghanistan proved (again) that even when you have overwhelming force you might not have a favorable outcome.
No, by far, China is the only nation which may ever consider taking the first strike.
There is Mao's famous saying, but also the fact China is super crazy about nuclear war preparedness.
Every major city has regular NBC drills, stockpiles of everything, and huge shelters.
Apple, Samsung, Intel, AMD, Texas Instruments, Nvidia, ASML, IBM, and so on. All are now welcome to take TSMC's intellectual property and use it in any way they see fit.
It'll be a spectacular race, and those companies have the resources to pursue it properly.
The same goes for all intellectual property in Taiwan. It all immediately becomes public property, free for anyone to utilize.
Who says? The US Government does, and the Europeans will likely agree. It'll usher in an epic confrontation politically and economically with China. The world will - practically overnight - split into two major factions, as with the Cold War.
Any people inside TSMC that are able to escape, or that are already in the US/Europe/etc, go to work helping the US and its allies to clone TSMC's technology. That includes spending the required sums to build fabs to accomodate that outcome. It'll be relatively easy for TSMC insiders to dump all relevant company trade secrets over to the US and its allies, which would be guaranteed to occur.
At $1B a pop + 50% for lifetime maintenance, the incentive is there.
Semi is heading to a winner take all market based on capital expenditure. Even TSMC can’t afford to in house the work ASML does and keep pace with new developments.