If you prefer to listen rather than read, this blog is available as a podcast here. Or if you want to listen to just this post:
I- Setting the scene
A few weeks ago, on October 7, the US imposed a new set of limits on exporting semiconductors (chips) to China. Not only that, but:
The new rules also restrict the ability of U.S. citizens to support the development or production of chips at Chinese fabs [fabricators] without a license.
This new set of rules attracted almost no attention until exactly a week later when those in the know suddenly started posting apocalyptic sounding tweets. (Okay, it probably wasn’t quite that sudden, but that’s what it felt like to me.)
This is what annihilation looks like: China’s semiconductor manufacturing industry was reduced to zero overnight. Complete collapse. No chance of survival.
(Which might not be an exaggeration, though it sounds like it must be.)
Despite these tweets, the story appears to still be mostly flying under the radar. Searching the daily news roundup emails I get from the NYT I found only one article about it, right next to a piece about the two girls who threw soup on the Van Gogh. And, actually, I’m positive that the soup throwing story has got more attention and is more widely known—particularly to the “man on the street”. I’m sure this won’t be the case forever, but it is an illustration of how important pivot points in history are often only obvious in retrospect.
I cannot foresee what the ultimate effect of this decision will be. we’ve just crossed a threshold, the kind historians decades from now will point to as one of those points of no return.
I don’t see a future now where the US-China relationship is not as dark as the Cold War rivalry.
So we have followed the thread back to its center and declared war on the future of China—its future military capacity, technological advance, and economic dynamism. We are now officially, openly, in the business of making sure China will not rise higher.
Is it really that bad? Have we crossed some sort of Rubicon as Greer maintains? Perhaps.
Experts are still trying to understand exactly what impact these new limits will have, but the US appears to have used its influence both domestically and with its allies to cut off China’s access to: advanced GPUs of the kind used for AI, the latest generation of advanced general purpose chips, and the equipment and software necessary to make such chips on their own.
Given that computer chips occupy a very special place in the structure of the modern world, and that China can’t construct such chips domestically, it’s very possible that none of the above was an exaggeration.
Something seemed to be brewing for a while on this front, and in order to understand it better I had started reading Chip War: The Quest to Dominate the World’s Most Critical Technology by Chris Miller, before this happened. It ended up being fortuitous, and as is so often the case, my good fortune is your good fortune, because I’m going to explain everything.
Actually, no, similar to The Princess Bride “…there is too much. Let me sum up.” And I’m probably not even going to be doing much summing up, that will come later when I do my formal review. What I’m really interested in is a brief survey of the potential futures if the US is going to try to cut China off from the “World’s Most Critical Technology”.
If you do want a good summation of things, and you don’t want to read a whole book. I would recommend this piece by Ben Thompson in his newsletter Stratechery. As for me I’m going to hit a bunch of points where I’ll largely be thinking out loud and asking questions:
II- Chips vs. Oil
Interestingly, only a couple of days before the US made its announcement about limits on exporting chips to China, OPEC+ announced that it was cutting oil production by two million barrels a day. This is another example of something that got far more attention than the chip restrictions, but in the long run will probably end up being less consequential.
The timing is interesting because one of the points that Miller makes over and over again in Chip War is that chips are as critical to an early 21st century nation as oil was to an early 19th century nation. And that in fact China spends more each year importing chips than it does importing oil. That will probably change with these restrictions. Which is a big deal both for China and for those nations and companies that are importing chips to China, a point we’ll return to.
Also, Miller points out that the bottlenecks in the chip industry are considerable:
Unlike oil, which can be bought from many countries, our production of computing power depends fundamentally on a series of choke points: tools, chemicals, and software that often are produced by a handful of companies—and sometimes only by one. No other facet of the economy is so dependent on so few firms. Chips from Taiwan provide 37 percent of the world’s new computing power each year. Two Korean companies produce 44 percent of the world’s memory chips. The Dutch company ASML builds 100 percent of the world’s extreme ultraviolet lithography machines, without which cutting-edge chips are simply impossible to make. OPEC’s 40 percent share of world oil production looks unimpressive by comparison.
You’ll notice that those are all essentially US allies, probably more tightly bound to the US than the nations of OPEC+ are bound to each other.
Speaking of oil, the nightmare analogy here is comparing Biden’s October 2022 restrictions on chips to Roosevelt’s July 1941 oil embargo against Japan. Roosevelt cut off 88% of Japan’s access to imported oil. Biden has (so far as I can tell) cut off 100% of China’s access to high end chips. In response to the oil embargo, Japan attacked Pearl Harbor. It remains to be seen what China will do in response to the chip restrictions. Though as it turns out the world’s premier supplier of high end chips, Taiwan Semiconductor Manufacturing Company (TSMC), is just 100 miles away from China, which brings us to:
III- China and Taiwan
TSMC produces 90% of the world’s most advanced chips, and the majority of production happens just across the strait from China in Taiwan. If there was some way for China to ensure that by capturing Taiwan they could seize that production, I think this would be the moment they tried it. Unfortunately for them, fabs are pretty fragile. The machines require unbelievable precision (which I’ll talk about more in a minute) and thousands of specialized parts. The fabs also require a staggering amount of expertise to operate, so the Chinese would not only have to figure out how to seize a fab without damaging it, but also figure out some way to secure the cooperation of the majority of the personnel. As both things are unlikely to happen, and given that China needs those chips, this setup has been called the silicon shield. China can’t invade Taiwan without destroying the supply of something it desperately needs in order to be the modern country it aspires to.
But, as Thompson points out in his newsletter, “one of the risks of cutting China off from TSMC is that the deterrent value of TSMC’s operations is diminished.” In other words if they can’t access TSMCs high end chips because of the recently passed restrictions then they suffer no harm from destroying those fabs, and there’s the opportunity to inflict significant harm on their global rivals.
Beyond all this I would add my normal caution that the shadowy future is the natural breeding ground of black swans and that the particular part of the future which lies at the intersection of China, Taiwan, the US and chips seems like exceptional fecund territory.
IV- China’s Path
In light of all of the foregoing, what will China do? What should China do? I haven’t seen any evidence that the chip restrictions are some kind of negotiating strategy. That the US is putting them out there as the first step in making a deal, i.e. “We’ll lift the restrictions on chips if you stop oppressing the Uyghurs, restore democratic norms in Hong Kong, renounce all claims to Taiwan, etc.” If such demands existed then China, and also we, as observers, could evaluate the ROI, but instead as Greer pointed out, we have “declared war on the future of China.”
Obviously, as has already been alluded to, it could escalate to actual war. But if we set that aside for the moment what other possibilities are there? The obvious course of action would be for China to build up their own ability to manufacture these chips. And certainly they’re doing this, but this is a profoundly difficult endeavor, which, even if everything goes perfectly, will take years to pull off. As Miller points out in Chip War:
China’s problem isn’t only in chip fabrication. In nearly every step of the process of producing semiconductors, China is staggeringly dependent on foreign technology, almost all of which is controlled by China’s geopolitical rivals—Taiwan, Japan, South Korea, or the United States. The software tools used to design chips are dominated by U.S. firms, while China has less than 1 percent of the global software tool market, according to data aggregated by scholars at Georgetown University’s Center for Security and Emerging Technology. When it comes to core intellectual property, the building blocks of transistor patterns from which many chips are designed, China’s market share is 2 percent; most of the rest is American or British. China supplies 4 percent of the world’s silicon wafers and other chipmaking materials; 1 percent of the tools used to fabricate chips; 5 percent of the market for chip designs. It has only a 7 percent market share in the business of fabricating chips. None of this fabrication capacity involves high-value, leading-edge technology.
No one doubts that China is willing to throw money and resources at the problem, the question is whether that’s enough. Can they be certain that given long enough they will eventually be able to make chips as good as TSMC? What is “long enough”? Is it reasonable to expect it in a year? Five years? Ten? Never? We will cover this last possibility in the next section, but it would seem that there’s also some time horizon past which it might as well be never. If it takes them 10 years to catch up, then at a minimum the state of the art is 10 years farther on. Plus everything that uses chips is 10 years behind. All of which is to say it’s hard to imagine that—if it were really going to take this long—that China would just sit back and decide to be patient.
On the other side of the equation, if it were merely 1 year, then they probably would try and wait it out. But no one thinks that China is one year behind TSMC. Intel is 4-5 years behind TSMC and everyone pretty much agrees that China would love to be where Intel is at the moment.
Now to be fair, China’s chief chip manufacturer, SMIC, has apparently produced some 7 nanometer chips, which is exactly where Intel is hoping to be at the start of next year. But can China scale this process up? It’s not enough to make one chip or even a thousand chips, you have to be able to make hundreds of thousands. TSMC is making 150,000 5 nm chips a month. Most people believe that even if SMIC managed to put together a 7 nm chip, it could not produce tens of thousands of them a month. Also it’s not clear if they can still make them after the imposition of the export limits.
To return to the question of “how long”, it’s hard to say, but if they really are cut off from high end Western chip technology, then I think the answer is longer than they are willing to wait. And maybe never, which takes us to our next section…
Out of all the components that go into making the most advanced chips, the scarcest and most expensive are the extreme ultraviolet lithography (EUV) machines. They are made by a single Dutch company, ASML. According to Chip War, the technology “took tens of billions of dollars and several decades to develop” and the resulting machines have “hundreds of thousands” of components from “thousands of suppliers”.
In the beginning a silicon chip might have had a few thousand transistors. But as Gordon Moore famously observed, the number of transistors on a chip doubled every two years. The technology for placing more and more transistors on a single chip involved using light of lower and lower wavelengths, thus the “extreme ultraviolet” designation. But as the wavelengths got lower the challenges mounted. Chip War gives an overview of the precision required:
If the mirrors in an EUV system were scaled to the size of Germany, the company said, their biggest irregularities would be a tenth of a millimeter. To direct EUV light with precision, they must be held perfectly still, requiring mechanics and sensors so exact that [the company] boasted they could be used to aim a laser to hit a golf ball as far away as the moon.
Before ASML finally put all the pieces together in 2018 the two biggest companies, Nikon and Canon, had given up. Beyond that several people publicly declared that it was impossible. One could make a credible claim that the EUV machines produced by ASML are the most technically advanced thing that humans have ever done. Each machine costs $150 million dollars and ASML has said that it’s not particularly worried about trade restrictions because their backlog is already so great.
It’s hard to know exactly how long it would take China to create their own EUV machines, but I’ve seen estimates of at least ten years, and this assumes that they throw everything they have at it: unlimited money, the smartest people, all the tech their spies can steal, etc. That sounds right to me, possibly even optimistic.
For the sake of argument let’s say that China is able to produce 7 nanometer chips at scale in a reasonable timeframe. Soon enough that they don’t end up doing something rash. Will it matter that EUV is still out of their reach?
Of the many battles being fought here, one of the biggest revolves around military capability. It’s possible that if AI becomes a big part of waging war, that the difference between an AI that can be built with TSMC/ASML 3 nm chips and the AI that can be built with Chinese 7 nm chips (assuming they can be built at scale) is like the difference between an F-22 and an F-15, or it could be that the difference is really not that great. Since military AI of the sort that will be used on autonomous drones is still in its infancy (and subject to numerous moral concerns) it’s not clear which it will be. But clearly if you could choose you’d want the more advanced chip.
The estimate of ten years could end up being never, given how much could change in that time. Also it’s possible that even if China had all the time in the world that they might never be able to create their own EUV machines. Certainly lots of countries are incapable of producing tech above a certain level. (Russia would have dearly loved to have its own high tech chip industry, but despite massive effort during the Cold War they just fell farther and farther behind.) Perhaps in order to get an ASML you need the innovation, liquidity, and drive of the entire western developed world. And China just can’t duplicate that by itself.
VI- The Economics of Things
We’re still at the very beginning of the US restrictions. Not only is it far too early to see how they’re going to play out, but also China has yet to retaliate with restrictions of their own—which they presumably will. In addition to waiting for China’s reaction there are also numerous other players whose reactions we’re still waiting on, collectively we call these players “The Market” and in the end their reactions might be far more decisive than China’s.
First let’s talk about where these two categories overlap. China is definitely part of “The Market”. Not only do they import a lot of chips, they import a lot of everything. They also export a lot of everything, including chips. We’ve been entirely focused on high-end chips, but what about low-end chips? China has traditionally done pretty well at competing at the lower end of things and chips are no exception. From Thompson’s Stratechery post:
In the end it was China that picked up a lot of the slack: the company’s [sic? I think this should be “country’s”] commitment to building its own semiconductor industry is not a new one (just much more pressing), and part of the process of walking the path I detailed above is building more basic chips using older technologies. China’s share of >45 nanometer chips was 23% in 2019, and probably over 35% today; its share of 28-45 nanometer chips was 19% in 2019 and is probably approaching 30% today. Moreover, these chips still make up most of the volume for the industry as a whole…
I’m not exactly sure how this will affect the “Chip War” but it at least demonstrates that China has some chip-making leverage of its own which it may use. We saw what happened during the pandemic when there were supply chain disruptions with chips. 35% and 30% are pretty big chunks of those supply chains should the Chinese choose to disrupt them.
Second, it should be noted that the US is mostly fighting this war through proxies. We’ve already mentioned that ASML and TSMC are foreign companies, but it goes deeper than just two companies, our own chip industry is in disarray. See for example this article from The Economist: The American chip industry’s $1.5trn meltdown, which points out that we’re suffering from a supply glut already. Or this article from Foreign Affairs: How Silicon Valley Lost the Chips Race. Meaning that while we don’t need ASML and TSMC as much as China does, we do need them, and it’s not clear how long they’re going to go along with the US throwing its weight around. Certainly they’ll play along for a while, but at some point one has to imagine that they’re going to run out of patience.
Of course there’s also the broader reaction companies and countries will have to these restrictions. We’ve recently seen pushback against similar attempts to shape international markets, for example the Saudis reneging on the secret deal the Biden White House thought they’d made to increase oil production. Chip restrictions would seem to fit into a similar narrative.
Obviously a full discussion of all the potential market distortions and other reactions these restrictions might engender is outside the scope of this post, but I would be very surprised if there weren’t large and unexpected negative second order effects from this policy.
VII- Longer term
As we saw in Part I, some people are framing this as a new cold war. When considering the long term implications of the war over chips, that’s probably a good way of looking at it, because it basically boils down to a competition between two systems. On the one side we have the globalized Western economies. On the other side we have a semi-planned, semi-communist, semi-facist authoritarian economy, which has been able to grow at a truly staggering rate over the last several decades.
Both systems are showing signs of weakness. As much as the Western world appears to be holding all the trump cards, they have historically been reluctant to really push that advantage. A trade war with China is going to be painful, particularly when the economy is already hurting. Globalism works best when it’s global, and China was already very tightly integrated with that system.
On the other side of things, China seems to be falling into the same trap most authoritarian regimes fall into, of centralizing power, which leads to myopia, which leads to bad policies. Consider China’s continued increasingly draconic and misguided efforts to maintain zero covid. Though it’s worth noting that the Chinese chip industry is so important that rather than shut it down, people are just living at the factory.
In general it boils down to a lot of things I’ve been talking about recently. Is Peter Zeihan right and China will wither if deprived of globalism? Is this the first step in that? Or is Ray Dalio right, and China is on its way up, and our chip restrictions will end up being too little too late? Both think that the Far East is going to end up being a very different place in the future. Zeihan thinks Japan will be the regional hegemon, Dalio thinks it’s going to be China. In either case an awful lot of the chip making capacity we depend on is in Taiwan, Korea and Japan. If the US loses its influence in the region then one could imagine the tables turning, with the US being the one cut off from high tech chips.
Francis Fukuyama, who’s enjoying a moment in the sun right now, would say that liberal democracies are better at science and science makes you better at war. And maybe it will come to that, and we’ll see which of the two systems is better. Right now China appears to be on the ropes, but I’d like to leave you with one final thought. TSMC is not on top because Taiwan is a model liberal democracy. TSMC is on top because the Taiwanese government gave them a huge amount of assistance, and US companies made some bad decisions. China is more than capable of providing similar levels of assistance, and we are still capable of making very bad decisions.
Now that I’ve shared my good fortune with you, it’s time for me to ask for you to share your good fortune with me. You know the drill…