Best of Q2 2024
War in Taiwan, AI, Japan, Predicting the News
We’ve had another great quarter of deep-dive content on the podcast and newsletter. The past few months have made clear how ChinaTalk is also very much in the business of nailing future headlines.
Last November, we ran epic breakdown of Commerce’s updated 2023 export controls as viewed by a thirty-year lithography veteran, predicting that BIS regulations would force ASML to stop shipping the 1970i and 1980Di to China. No surprise when BIS blocked precisely those machines a few weeks ago.
As the PLA purges were just getting underway, we brought on Joel Wunthow to tell us just how bad Xi’s trust issues with the PLA are, and what that means for China’s overall military readiness. No surprise when Bloomberg reported at the top of 2024 that Xi’s purges were prompted by military corruption so pervasive that (per US intelligence assessments) Xi is now “less likely to contemplate major military action in the coming years than would otherwise have been the case.”
And ChinaTalk’s latest prediction comes from Irene Zhang’s excellent investigative reporting on how easy it is to buy advanced chips on Xiaohongshu, China’s Instagram. No surprise when — a full year after Irene’s report — the WSJ covered the retail smuggling ecosystem, including one student in Singapore who smuggled six Nvidia semis back to China (for a whopping 0.4% commission).
Subscribe for free to get breaking news in your inbox months before they’re mainstream talking points.
Now, on to Q2 hits.
War in Taiwan Month
It’s the question we get asked all the time: “Will China invade Taiwan?” We had some of our best discussions of Q2 taking a stab at answering it.
First, we hosted Jared McKinney of the Air War College and Peter Harris of Colorado State University, who recently co-authored a fantastic monograph, “Deterrence Gap: Avoiding War in the Taiwan Strait.”
Their major value-add to the Taiwan discussion: interlocking deterrents (as opposed to deterrence). The reason cross-Strait conflict has been held at bay for so long, they argue, is because the PLA was simultaneously facing several deterrents, including US naval dominance, Taiwan air superiority, and even the US threatening to use nuclear weapons.
Here’s why there’s cause for concern today:
Constraints (ie. externally imposed limitations) against the PLA have all but disappeared, and restraints (ie. internally motivated reasons to exercise self-control) are quickly dissipating, too.
But there’s a way out. Here’s how McKinney and Harris propose we start layering up deterrents again — and it’s actually not that expensive:
Is there any precedent to invading Taiwan? US Army Lieutenant Colonel J. Kevin McKittrick tells us there is: in 1944, US generals set their sights on Japanese-controlled Formosa as a staging area for attacking Japan and ending WWII.
American warplanners ultimately aborted their Taiwan-invasion plans, however, because the resource constraints of this massive amphibious invasion were just too taxing. Conventional army doctrine has it that the attacker-to-defender ratio should be 3:1 for >50% odds at success; for more complex operations — like amphibious assaults — the ratio should probably be up to 5:1.
The same resource constraints that existed for the US in 1944 probably still exist for the PLA today. LTC McKittrick:
The challenge is that the number of troops that you need to land exceeds the capacity of the Chinese Navy — even with the roll-on, roll-off civilian vessels — to get that number of troops on the ground at the same time.
We just brought up the figure that came from Thomas Shugart’s testimony before the US-China Economic and Security Review Commission, where he stated that it would take roughly 10 days to deliver approximately 300,000 troops on the ground.
That probably is not enough troops, and that is probably too slow.
If Taiwan, conservatively, has 244,000 troops ready to go at a moment’s notice — which it probably does — then using a 3:1 force-ratio requirement, the PLA would have to send 732,000 troops across the Strait. If they want a 5:1 option, the PLA would need 1.2 million troops. It would take more than a month to move that many troops across the Strait.
So, would Xi still send it? Dmitri Alperovitch thinks it’s possible:
It’s true that Putin is much more of a gambler than Xi — I agree with you on that. But I still think that Xi is a gambler.
Xi has gambled and engaged in adventurism, whether it’s the Sierra Madre in the South China Sea, or whether it’s the confrontation with India on the border. Or we can look at zero Covid, which was the ultimate gamble — he shut down the whole country for several years in the hopes of stopping a virus that spreads through air droplets.
Xi clearly wants to invade Taiwan, or at least have that option. You see that from the intelligence community assessments, but you can also see it because the military buildup that he is pursuing is specifically focused on that one thing.
But there’s reason for hope — “we can survive if we settle the Taiwan issue.” More from Dmitri:
I actually think China is a lot weaker than the Soviet Union was during the Cold War, and we beat them. For sure, we can beat the Chinese and make sure that this century remains an American century. But the first thing we have to do is avoid a conflict and hostile takeover of Taiwan.
Top Pods: Scale AI, Sen. Young, India-China, TSMC in Arizona, Shakespeare and Power
Scale AI CEO Alex Wang: ChinaTalk has done a ton of coverage on compute over the past few years — but Alex was the first “data” guest of the pod. “For a lot of the problems that we want these models to solve,” Alex said, “the data is not recorded or does not exist.” That’s why we need “frontier data” — high-quality data like logical reasoning, decision matrices of subject-matter experts, complex math — both for successful innovation as well as for national-security imperatives:
It’s quite frustrating. … For less nebulous technologies, we usually wait until we’ve been surpassed before we get our act together. You can see this with hypersonic missiles, as well as a lot of developments in space technology — we’ve waited for China to catch up a little bit before reigniting the flames. I have a real concern that it will take a crisis to make us take this stuff seriously.
Senator Todd Young (R-Ind.): One of the few elected legislators who actually legislates, Sen. Young was the latest member of Congress to hop on the pod and discuss tech-policy futures. The CHIPS Act co-author wants to beef up Congress’s tech smarts, revamp the civil service to lure in AI talent, and unlock more data for AI development (without forgetting about property rights). He’s bullish on R&D funding and high-skilled immigration, and he’s even a fan of the AI Safety Institute (don’t mind the GOP platform — “I also know that sometimes these things are developed by 22-year-old interns”).
C. Raja Mohan (ft. James Crabtree): Modi’s on a roll. He recently clinched a third term, and he’s wooing Western capital like it’s going out of style. As paraphrased by Raja, Modi is basically saying, “What you’ve seen in the last 10 years is just a trailer. The real movie will begin now.” See, e.g., three semiconductor plants, and a potential Elon trip to India later this year. But it’s not just about economic gains — it’s also a strategic pivot to a coalitional strategy:
We are in a Dengist moment in India. The parallel extends to India’s growth, and India’s opportunity to work with American capital to build itself up. Across the board, I would say, it is an era of partnership with the US. The fact is, the differences between India and China are huge, and the disputes have become bigger.
[China’s] assumptions that their neighbors were too weak to come together and that America was in decline both turned out to be wrong. … [Crabtee:] It turned out India actually did have some options. It just had to junk many decades’ worth of ideological baggage, which it has done with great speed.
Viola Zhou of Rest of World: TSMC Phoenix has a culture clash on its hands. Taiwanese engineers are scratching their heads at yanks who dare to eat during meetings and clock out at 5 p.m. sharp, while American workers are balking at the militaristic management style (indeed, Taiwan-based TSMC employees are used to calling their managers zhǎngguān 長官, a military term that means “commanding officer”). We even have a TSMC employee on record saying, “I want to support TSMC to be great. It’s my religion.” And here’s a nominee for quote of the year (oh how I miss Taiwan):
The bustling cities of Taiwan are densely packed and offer extensive public transport, ubiquitous street food, and 24-hour convenience stores every few blocks. In northern Phoenix, everyday life is impossible without a car, and East Asian faces are scarce. “Everything is so big in America,” said one engineer, recalling his first impression. He recounted his wife summarizing her impression of the US: “Great mountains, great rivers, and great boredom.”
Eliot Cohen (ft. Phil Schneider) — part 1 and part 2 (ft. Drew Lichtenberg and Kate Pitt): Shakespeare has more to teach us about power than any political science textbook. From the St. Crispin’s Day Speech of Henry V to Mark Antony whipping up the crowd, the Bard nails how leaders manipulate the masses. But he also conveys a real empathy for the burdens of the crown — consider Prospero tossing his magic wand, or Richard II’s impaired sense of reality even after he’s deposed and in prison. Turns out, knowing when to exit stage left might be the trickiest act of all. Biden and Trump, take note: “There is a world elsewhere” beyond the Oval Office. Special thanks to Jordan’s brother for bringing Shakespeare to life in these episodes.
Meanwhile in the Mainland
Welcome aboard to two new ChinaTalk contributors: Bit Wise (pseudonym) and Yiwen Lu.
Anyone can talk all day about what they think China should do about its various tech bottlenecks and structural industrial challenges, but Bit Wise brings us coverage of what the top dogs in China are actually saying among themselves. Three AI heavyweights — Xue Lan 薛澜, Zhang Hongjiang 张宏江, and Li Hang 李航 — speak openly in a panel discussion about the questions we’ve all been wondering about:
Does Sora understand physics? (If so, is a robotics revolution around the corner?)
Talent, data, compute — what’s the real bottleneck in China?
Which business strategies lend themselves better to open-sourcing models?
Should frontier models be open-sourced? And should Chinese companies even be focusing on frontier models?
China’s hottest model, SenseChat V5, was put through the wringer by Bit Wise and
. They compare SenseTime’s performance markers to publicly available rankings (and report on SenseTime’s personal response to their inquiries), run some censorship tests, and also feed some excellent memes to the chatbot. And of course, Lily “math is for everyone” Ottinger takes SenseChat to town on a college-level calculus exam.In late June, China lost API access to OpenAI’s services — a decision that signals a tightening of AI export controls, driven by cybersecurity concerns and fears of model extraction. Yiwen’s piece, though, investigates the extent this decision was really anything new; after all, OpenAI and Chinese regulators have each put up their fair share of barriers already. Also, why is OpenAI leaving China when others — like Microsoft’s Azure — are digging in even further?
In some ways, this is probably the puzzle for all big enterprises: despite all the heartache, it still doesn’t make sense to give up on China. Apple, which just had a grand unveiling of its partnership with ChatGPT to develop Apple Intelligence for its products in the US, reportedly held preliminary talks with Baidu about using the latter’s AI technology in its devices in China — precisely because its flagship global partner ChatGPT is not available in China. Similarly, Samsung uses Baidu’s ERNIE 文心一言 within its latest smartphone model in China, and Google’s Gemini outside of China.
Also: is China becoming a “great nation”? Zhang Yunchen 张运成 of CICIR — a think tank nestled under China’s Ministry of State Security — thinks so. If you want a window into what the CCP’s top leaders are hearing from their advisors, this translation by Dylan Levi King and Nancy Yu is a must-read. The cliffnotes:
Competitions between nations are won by superior resources, territorial advantages, and vast populations.
But scientific, technological, commercial, and political advancement can compensate for shortfalls in resources, territory, and population.
China, Zhang says, rose in the late twentieth century through territory and population — but it has continued on the road to becoming a great nation because it has seized scientific, technological, commercial, and political advances.
US Competition Policy
$11 billion of CHIPS Act funding is headed for the US National Science and Technology Council (NSTC). What should NSTC do with those funds? Jordan and Chip Capitols founder Arrian Ebrahimi make the case in a long-form report that corporate supplements to existing R&D just aren’t going to cut it anymore. We need moonshots. Here’s the TL;DR game plan they propose (9,000 → 160 words):
Research Agenda: Focus on addressing market failures and underinvested areas in semiconductor research. Prioritize proactive design for manufacturing, de-risking new materials and eco-friendly chemicals, and organizing cross-sectoral missions for downstream technologies like AI, security, and sustainability.
Investment Strategy: Focus on Series A and B startups to bridge the “valley of death” and support ventures outside current trends. Act as a lead investor to organize funding syndicates, leveraging technical expertise and financial resources.
Intellectual Property Sharing Model: As the consortium matures, implement a dynamic approach evolving from open access to more restricted models. Tailor IP rights to match members’ contributions and research stages.
Geographic Structure: Establish a centralized flagship campus with strategically placed satellite “Centers of Excellence.” Ensure all facilities support centrally determined technology goals rather than pursuing independent verticals.
Funding Trajectory: Follow a three-phase approach over 10+ years, transitioning from primarily public funds to a balanced structure of participation fees and fee-for-service projects. Maintain public funding for infrastructure expansion throughout.
If we play our cards right, the NSTC could be the secret sauce that keeps America at the top of the chip game. It’s not just about making faster phones — the US could achieve breakthroughs in national security, economic dominance, and maybe even save the planet.
Next up: we produced two back-to-back deep dives into the Biden administration’s tariffs on Chinese EVs:
Brad Setser — formerly in Biden’s USTR — hopped on an emergency pod to discuss the national-security justifications for the EV tariffs, why the Biden administration waited so long before imposing these tariffs, and how the policy response to today’s Chinese EV manufacturers differs from the response to Japanese automakers in the 1980s.
But the wrinkle, per Kevin Zhang: overall, Chinese EVs are just better than their Western rivals, winning in price competitiveness and sporting fancy features that make Tesla look like a Nokia brick phone. So the Biden administration is betting big, trading short-term consumer choice for long-term industrial might. TBD whether the “electric curtain” will spark an American EV renaissance, or leave US drivers in the dust while the rest of the world picks up Chinese EVs anyways.
How do these kinds of policy choices play into a coherent China strategy? (Does the US even have a coherent China strategy?) Do check out our rapid response to the latest mashup at Foreign Affairs, featuring many friends of the pod: Matt Pottinger, Mike Gallagher, Jessica Chen Weiss — plus James Steinberg, Paul Heer, and Rush Doshi.
Japan
ChinaTalk’s first ever staff retreat in Tokyo also let us harvest some excellent content. Jordan published one extended meditation on Japan and three Japan podcasts:
Ambassador Rahm Emanuel: This one is a classic. Rahm doesn’t mince words — he articulated the Biden administration’s view of China today with language that nobody in the Biden administration would use themselves.
China has quite an entrepreneurial culture, and [Xi] crushed their entrepreneurship — and in crushing their entrepreneurship and the tactics he adopted, he crushed the confidence of the world in China. … You can’t get anybody in Japan, Europe, or the United States to raise their hand and say, “I’d like to move my family to a city where I could get arrested any given day and be in lockdown.”
Arrian and Jordan brought on University of Tokyo professor Kazuto Suzuki, who serves as an advisor to Japan’s Ministry of the Economy, Trade, and Industry (METI). Since a tense island dispute with China in 2010, Japan has been busy fortifying its supply chains and pouring yen into R&D — so METI is back in the spotlight, dishing out billion-dollar checks to semiconductor projects, juggling partnerships with frenemies like South Korea, and rewriting the export-control playbook. And as METI charts this course, it’s drawing lessons from its past — from Qing Dynasty tech woes to WWII trade conflicts — to avoid repeating old missteps.
ChinaTalk next hosted Ryan Takeshita, Chief Global Editor at PIVOT, a new media outlet in Japan focused on the emerging startup scene. In that conversation, we get into the recent economic history leading to today’s “boom times,” why more people are looking to leave traditional occupations for insurgent firms, and challenges surrounding demographics and immigration.
And of course, we have the travelog report from Jordan — from rubbing elbows with METI officials to chowing down on Kyoto’s finest. He also took up some Zen Buddhism (spoiler: it’s complicated), hung out with samurai ghosts, and even uncovered an Art Deco palace with a dark past. Between those adventures, Jordan found time to read classic Japanese literature, contemplate the merits of batting cages over meditation, and stumble upon Japan’s celebrity-endorsed mushroom market.
Media Diet
Jake Newby of Concrete Avalanche: Best Chinese music of 2024? We covered everything from psychedelic rock and Beijing kawaii core, to rare Uyghur folk and Tibetan Buddhist chants. One listener said, “This deradicalized me a bit — I didn’t know people made music in China like this!”
If you liked that pod, you’ll like ChinaTalk’s recent hire, Alexa Pan, whose first piece asked, “How would William Wordsworth’s The Prelude work with Chinese shoegaze?”
H1 2024 reading list from Jordan:
Mitsui Empire: Three Centuries of Japanese Business, by John Roberts — the best political economy book on Japan.
Ed Luttwak’s works: Strategy: The Logic of War and Peace; The Grand Strategy of the Byzantine Empire; and The Endangered American Dream — the last of which was quite prescient on the importance of domestic manufacturing.
History of the US Civil Service, by Paul P. Van Riper — there should be more administrative-state histories!
The Ties That Bound: Peasant Families in Medieval England, by Barbara A. Hanawalt — takeaway: life totally sucked until recently.
And finally: the betting markets.
Will OpenAI, Google, or Anthropic have a model’s weights hacked before 2025?
Jordan set the market at 30%.
Market at Q1: 27%.
Market today: 13%.
That said, we did get an OpenAI hack story and Gen. Nakasone appointed to the board! This feels like a buy to me…
Jordan set the market at 15%.
Market at Q1: 35%.
Market today: 41%.
Will China get mentioned more times than AI over the course of the 2024 presidential debates?
Jordan set the market at 60%.
Market at Q1: 91%.
Market today: 89%. (Check out Lily’s Biden-Trump debate roundup!)
Will a SMIC 5-nm chip make it into production in a Huawei device in 2024?
Jordan set the market at 65%.
Market at Q1: 27%
Market today: 59%.
Will a Politburo Standing Committee or State Council member get purged in 2024?
Jordan set the market at 25%.
Market at Q1: 15%.
Market today: 15%.
Thanks for sticking with us. Please consider a paid subscription.
Thank you Nicholas for an very interesting post with unusually diverse range of readings and links - I'm still enjoying following them all up. A great 'best of' indeed.