My last reading series was about China, so I decided to make my next one about a related topic: U.S.-China technological competition. This is of huge economic importance, because government policies aimed at dominating strategic high-tech industries are poised to reshape the global economy. It’s also interesting in its own right, because now we’re getting to see the big innovations of the past few decades in a very different light. I’ve been predicting for a while that we’d see a rotation from civilian applications of information technology to military or quasi-military ones, and now we’re seeing that happening.
There are a number of books on this subject that seem more general in scope (The Wires of War, The Kill Chain, and Digital Silk Road, to name just a few), and I’ll get around to reading some of these. But for my first foray, I wanted to focus on three specific technologies that seem like they’re at the heart of the superpower rivalry: semiconductors, wireless networking, and AI. I’m basically working off of the Special Competitive Studies Project’s list of “key technologies”:
All of these are important, of course, but chips, wireless, and AI seem like they’re the items that are:
currently widely deployed (as opposed to quantum computing, fusion, etc.)
of immediate and obvious military application, and
rely heavily on cutting-edge research.
I think it’s no coincidence that these are the three areas that have been most hotly contested in recent years. So I picked one book to read about the struggle to dominate each of these technologies.
1. “Chip War”, by Chris Miller
This was really the standout book of the bunch; it appears on many lists of “best books of 2022”, and the acclaim is well-deserved. Miller offers a real master class on how to write an accessible, engaging book about an esoteric technical subject. He starts with the narrative of how the U.S. invented the semiconductor industry, and how it defended its supremacy in that industry from challenges by the USSR and (especially) Japan. He then moves on to how the U.S. outsourced much of the industry to Taiwan, South Korea, and the Netherlands during the age of globalization, before finally discussing China’s challenge and the U.S. response in the 2010s.
All along the way, Miller explains the intricacies of chipmaking technology in surprisingly clear and simple terms, from the days when silicon wafers were cut by hand, to the days when cutting-edge chipmaking requires the world’s most powerful lasers and the world’s smoothest mirrors. At the end of this book, if you didn’t know already, you’ll understand the difference between the capital requirements of the NAND flash memory industry and the DRAM industry, between DUV and EUV lithography, and between a foundry and a fabless design firm. And you’ll have enjoyed learning about them, because they were interwoven seamlessly with stories about the colorful heroes of the chip industry, like Bob Noyce, Morris Chang, and Andy Grove.
My favorite thing about Miller’s book, though, is the relentless focus on why the chip industry is so strategically important. As I wrote in a post last year, it’s really all about precision weaponry. The pinpoint HIMARS strikes that are devastating Russian forces in Ukraine, the precision bombs that decimated Iraqi forces in 1991, and the long-range missiles that China is building for a possible war over Taiwan all rely crucially on advanced computer chips to tell them where the enemy is. Semiconductors mean death on delivery. And in the age of AI this will only be more true.
Anyway, to make a long story short, the U.S. is winning the semiconductor wars for now. Despite the fragmentation of the supply chain for the highest-end chips, all the basic components are in the hands of either the U.S. or its allies.
The U.S. makes the design software and the high-end lasers for lithography, Germany makes the lithography mirrors, the U.S., UK, and South Korea design the chips, the Netherlands and Japan make the lithography machines, and Taiwan and South Korea actually fabricate the chips. There is danger in this, because China could blockade, bombard, or conquer Taiwan and South Korea pretty easily in a war situation, and because Taiwan especially is threatened by earthquakes. But overall we’re not reliant on China for any part of this industry.
And thanks to export controls, we likely won’t be. Over the last few years, the U.S. has been rolling out increasingly sweeping and stringent restrictions on the sale of chips and (more importantly) chipmaking equipment to China. These controls require international coordination that’s sometimes difficult to enforce, but so far they seem to be highly effective. Though many predicted that export controls would create a “Sputnik moment” that would prompt China to zoom ahead in chip technology, so far its efforts have fallen flat in a big way.
That doesn’t mean China won’t have enough chips to make cutting-edge precision weapons, of course. As Miller explains, the country will still be able to make vast quantities of “trailing-edge” chips, which may be good enough to win in a conflict. But for now at least, the U.S. looks like it has a decisive advantage in the new technological cold war.
2. “Wireless Wars”, by Jonathan Pelson
If semiconductors are the area where the U.S. has the upper hand, wireless tech is the area where China has kicked our butt so far. In Wireless Wars, former telecom executive Jonathan Pelson tells the story of how this happened, and offers some suggestions for how the U.S. might reclaim the upper hand.
Unlike Chip War, Wireless Wars can be a frustrating book to read. Pelson spends a lot of time narrating telecom industry war stories; this makes sense, since he was on the ground for a lot of those battles. But because he doesn’t accompany these stories with a general explanation of how the telecom industry works, scenes that are written in a breathless tone aren’t as gripping as the author wants them to be; most readers just won’t know the stakes.
The basic idea, I think, is that the telecom industry is defined by network effects — instead of a bunch of companies taking little bits of the market, all the companies are constantly engaged in a desperate battle for absolute supremacy. Top-quality technology is necessary but not sufficient; you have to be able to buy your competitors or drive them out of business. In the 90s, after Bell’s comfy monopoly had been shattered by a combination of antitrust and new technology, the industry was defined by tons of American companies using the cheap capital of the dotcom bubble to try to dominate the wireless equipment industry. They pretty much all failed and died. European companies sort of took over by default, but soon everyone was driven to the margins of the market by the new champion: China’s Huawei.
Wireless Wars is basically the story of Huawei. China’s current global dominance in wireless equipment is basically a function of this one company. Huawei combines several major advantages, both fair and unfair — a cutthroat corporate culture, a world-class research division, rampant IP theft, deep ties with the Chinese security services, and state subsidies that give it much deeper pockets than its rivals. Essentially, Huawei either steals or invents top-line tech, then goes around offering its equipment for such a cheap price that there’s really no way for it to make a profit. Even when its European rivals manage to win contracts (usually based on national security concerns), Huawei’s ultra-cheap prices force those rivals to cut their own prices, robbing them of the profits they need to conduct R&D to keep up with Huawei. As a result, despite bans in a number of countries, Huawei is poised to be the telecom provider for much of the developing world.
This is strategically important because it’ll allow Huawei — and thus, China’s spy agencies — to spy on everybody who uses its equipment; when the U.S. banned the company, it was in the process of trying to establish outposts near nuclear weapons sites.
That’s basically the story as Pelson tells it, and he takes care to issue lots of caveats and mention alternative interpretations. But whether all the details are exactly right, the overall story is quite plausible and fits closely with other things I’ve read. The most astonishing and disturbing part of the book is when Paulson lists a bunch of high-level Western figures who warned strongly against allowing Huawei, but who were later hired as Huawei spokespeople and changed their stance completely. The ease with which these people, mostly in the UK, were essentially bought by a foreign government underscores what a disadvantage free societies have in this kind of fight.
At the end of the book, Pelson has a suggestion about how free societies can play to their own strengths in order to fight back. The other frustrating thing about Wireless Wars is that Pelson rarely explains how wireless technology actually works — but at the end, he remedies this somewhat. He explains that in the past, the boxes of equipment in wireless towers contained components and software that were made by only one company, which is why the industry tended to coalesce around one dominant player. But new technologies would disaggregate the functions of these boxes, allowing wireless carriers to buy the different pieces from a whole bunch of different vendors — as long as regulators allow this. Switching wireless tech from a closed to an open format, Pelson argues, would kill much of the telecom industry’s fabled network effect, allowing providers to focus on narrow high-tech areas of competency and theoretically making it much harder for Huawei to offer a full package deal. The strategy sounds promising, and seems better than trying to prod Nokia or Ericsson to go toe-to-toe with the Chinese champion.
3. “AI Superpowers”, by Kai-Fu Lee
This book is much less directly useful than the other two; the author explicitly states that the struggle for AI dominance between the U.S. and China will be about peaceful commerce rather than war, and studiously avoids any discussion of military applications. That attitude is a reflection of the fact that it was written in 2018; what a difference a few years have made. But in any case, AI Superpowers is worth a read because it illustrates exactly how China’s boosters (of which Lee, who now lives in Beijing, is definitely one) believe the country will vanquish the U.S. in the battle for AI supremacy.
Lee’s main thesis is that AI has essentially matured as a technology; the field’s major discoveries, he argues, were complete by the mid-2000s, and AI is now mostly about incremental innovations. That shift, he argues, means that China will be able to dominate the U.S. through a combination of plentiful data, ruthless entrepreneurship, engineering talent, and government support. The data comes from China’s gig economy and internet-of-things companies, and from state surveillance (though Lee mentions this only briefly). Lee describes China’s entrepreneurs as hard-working hyper-competitive copycats who will be able to imitate and then outdo Western rivals who are more focused on originality. The first two thirds of the book are basically Lee going through a bunch of functions and applications of AI technology, and issuing a prediction that China will eventually come to dominate each one.
Four years later, it’s interesting to see how wrong most of Lee’s predictions were. The invention of transformers and diffusers, which have sparked a revolution in generative AI, put paid to his thesis that AI has no more room for big discoveries. Gig economy services like Uber have proven underwhelming in the U.S. consumer space even without any Chinese competition, raising the question of whether China’s embrace of these services was largely due to its lower income levels. AI-powered innovations like the ones Lee predicts — basically, variants on Alexa and smart homes/cities — have so far not been very successful in the market. China’s government support for hard-charging entrepreneurs in the information technology space has proven fickle; the companies Lee extols the most were the subject of a harsh crackdown in 2021, after Xi Jinping decided they weren’t the most important areas of technology after all.
As for those talented AI researchers, it turns out that China does produce plenty of them, but most of them want to work in the U.S.
Anyway, none of this tells us the answer to the question of which country is actually leading in AI technology. In fact, that question is extremely hard to answer, because unlike in semiconductors (where you can measure chip capabilities) and wireless (where you can measure market share), it’s not clear what “leading in AI” actually means. Will OpenAI’s large language models be the most important machine learning application going forward? Or will more traditional “predictive” AI be more important because that’s all you really need in order to land a missile on its target? Even if we decide it’s the latter, it’s hard to tell which country is ahead, because these capabilities are largely classified.
In fact, there was recently an interesting — and inconclusive — debate about whether China has surpassed the U.S. in AI research. Nikkei Asia recently released a report claiming that China was ahead of the U.S. in both quantity and quality of papers published in the field. But Paul Scharre pointed out that in AI conference and repository citations, which may be a better barometer of quality, the U.S. still maintains a comfortable lead:
To me, this debate shows just how difficult it is to assess national leadership and capabilities in AI — really, the technology is just too new to make that kind of judgement. We don’t really know exactly how AI will be used on the battlefield of a future war, and thus we don’t know which measurements of AI capabilities are most important for assessing national power.
So while the U.S. leads in semiconductors and tries to catch up in wireless, in the third big contested area, there’s little clarity about where the competition stands.
I had got halfway through the Kill Chain a couple years back, and decided to pick it up again after the war in Ukraine started. It's fascinating just how much it seems to be written in a different world. Obviously the goal of the book was to jolt the US establishment out of complacency, so it presented some best-case scenarios for the Russian and Chinese militaries, but man comparing the Russian Army of that book to the one we saw in real life is extraordinary. A good reminder of how little we often know about others' capabilities, or our own as well - and how the error could really go in either direction.
"To me, this debate shows just how difficult it is to assess national leadership and capabilities in AI — really, the technology is just too new to make that kind of judgement. We don’t really know exactly how AI will be used on the battlefield of a future war, and thus we don’t know which measurements of AI capabilities are most important for assessing national power."
The fact that the U.S. government has let Taiwan, a small island nation of twenty four million people off the coast of China, a hostile power of 1.4 billion people, become the world's epicenter of microchip technology is shameful and a national security risk. We see the military risks of letting microchips be beholden to foreign powers, see Russia's desperate attempts to reclaim microchips from appliances to build military weapons systems in light of the sanctions, as a strategic risk.
All current AI or ML technologies are running on specialized microchips in the form of GPUs, of which the preponderance come from Taiwan, whether NVIDIA or AMD. On-shoring the manufacture of these designs, be it in Intel fabs in Arizona or the proposed fab in Ohio, should be of primary concern. The fact that the U.S. is spending $858,000,000,000 in the latest military budget yet is unconcerned about the sourcing of these processors is disheartening at best.