Xi Jinping’s paranoid approach to AGI, debt crisis, & Politburo politics — Victor Shih

Summary

Victor Shih, director of the 21st Century China Center at UC San Diego, sits down with Dwarkesh Patel for a 90-minute deep dive into how the Chinese Communist Party actually works — the day-to-day life of Xi Jinping, the structure of the leading small groups, who really controls AI, why local government debt has crossed 120% of GDP, and what happens when Xi is no longer there. Shih walks through the unusual ascent of Ding Xuexiang — the metallurgy-trained princeling who manages every flow of information across Xi’s desk and now runs both the Central Cybersecurity Commission office and the Central Science and Technology Commission. Ding’s job, Shih argues, is fundamentally not about accelerating AGI but about being able to “pull the brakes” on it: the party’s core anxiety is that AI becomes a tool used outside or inside the party to usurp its power.

The middle of the conversation is a clinic on Chinese political economy. Shih lays out why local debt has ballooned even as the central government brags about a 60-70% sovereign debt ratio: roughly 120-140% of GDP in additional debt sits at the local level, structured as authorized issuance the central government tells provinces to do without funding. He explains why “socialist banking + capitalist firms” produces structural overinvestment, why for every successful state-financed project there are a dozen failed ones, why semiconductor approval officials had to be jailed by the dozens, and why the rent-seeking incentive on infrastructure is what drives China’s overbuild — not GDP-promotion targets. He also lays out the four-stage Stalin/Xi parallel and identifies Bo Xilai as China’s “Trotsky” figure — the flamboyant, articulate competitor who never had a chance.

The final stretch covers succession, Taiwan, and macroeconomic forecasting. Shih thinks there is no succession plan — if Xi died tomorrow, the financial system would seize within weeks because the bureaucratic discipline that holds capital controls together depends entirely on Beijing watching. He sees Bo Xilai’s son in Canada as the most interesting potential interview Dwarkesh could do. On Taiwan, he reads Xi’s revealed preference as “wants it but not enough to bet everything” and notes that Putin’s failure in Ukraine likely raised Xi’s invasion threshold. His base-case 2040 forecast: China’s economy is roughly the same size as the US’s, not the dominant force most extrapolations imply. Shih’s bearish read on China’s institutional capacity stands in striking parallel to Ken Rogoff’s bearish economic forecast in the same channel’s catalog — a recurring 2025 Dwarkesh theme.

Highlights

”If Xi expresses a preference, you go that direction or you get purged”

“If Xi Jinping expresses a preference over a policy direction, that is a direction you have to go toward no matter what. Because if he observes that you’re dragging your feet or that you’re pursuing a slightly different agenda, you will be purged.” — Victor Shih, 12:00

yt-dlp --download-sections "*12:00-13:00" "https://www.youtube.com/watch?v=b1TeeIG6Uaw" --force-keyframes-at-cuts --merge-output-format mp4 -o "xi-purges-feet-draggers.mp4"

”There’s a team in every internet company who can pull the plug”

“There’s such a team of people in every major internet company in China… if AGI started generating Falun Gong related content and it proliferates very quickly beyond the capacity of the censor to control it, then they’ll have to stop the algorithm.” — Victor Shih, 30:00

yt-dlp --download-sections "*30:00-31:00" "https://www.youtube.com/watch?v=b1TeeIG6Uaw" --force-keyframes-at-cuts --merge-output-format mp4 -o "pull-the-plug.mp4"

”The He Lifeng story: he built the New Manhattan in Tianjin and then was promoted”

“The chief negotiator with the United States, He Lifeng, vice premier in charge of finance. He is well-known for starting and perpetuating the largest real estate bubble the world has ever seen in Tianjin… There’s the ‘New Manhattan,’ literally the size of Manhattan, filled with empty office buildings.” — Victor Shih, 33:00

yt-dlp --download-sections "*33:00-34:10" "https://www.youtube.com/watch?v=b1TeeIG6Uaw" --force-keyframes-at-cuts --merge-output-format mp4 -o "new-manhattan-tianjin.mp4"

”The brakes matter as much as the engine — China’s posture toward AGI”

“We need to invest in AI, but we need to do it — we can’t go all out in investing in it without knowing what the brakes are. We have to develop the brakes also at the same time.” — Victor Shih, paraphrasing Ding Xuexiang at Davos, 25:46

yt-dlp --download-sections "*25:46-27:00" "https://www.youtube.com/watch?v=b1TeeIG6Uaw" --force-keyframes-at-cuts --merge-output-format mp4 -o "agi-brakes.mp4"

”If Xi died tomorrow, China would have a financial crisis within two weeks”

“If there’s any sense that there’s no command structure in Beijing even for a week or two, I think we would have a financial crisis at least… it doesn’t take a panic. Even if people want to just rationally reallocate 10% of their assets outside of China, foreign exchange is gone, massive devaluation.” — Victor Shih, 1:21:00

yt-dlp --download-sections "*1:20:00-1:21:45" "https://www.youtube.com/watch?v=b1TeeIG6Uaw" --force-keyframes-at-cuts --merge-output-format mp4 -o "xi-dies-financial-crisis.mp4"

”Why DeepSeek beats American models for understanding the CCP”

“The most helpful has been DeepSeek… DeepSeek, which of course was first developed by a hedge fund to trade in the Chinese market, part of what they really trained the model for was to detect important policy documents and meetings within the Chinese government.” — Victor Shih, 36:39

yt-dlp --download-sections "*36:39-38:00" "https://www.youtube.com/watch?v=b1TeeIG6Uaw" --force-keyframes-at-cuts --merge-output-format mp4 -o "deepseek-research-tool.mp4"

Key Points

• China is highly fiscally centralized despite its surface decentralization (1:46) - The 1994 tax reform clawed back VAT to the center; locals get reimbursed conditionally
• Politburo technocrats are real but heterogeneous (4:32) - Ma Xingrui (Tsinghua), Zhang Guoqing (military industry), Ding Xuexiang (metallurgical forging) — competence varies hugely
• Mandatory ideology classes, optional economics (10:21) - STEM-track Chinese students never have to learn supply and demand; they DO have to take a “Situation and Policy” course annually
• Xi spends 270 days a year in policy meetings (13:13) - Genuinely substantive, with off-script Q&A; his post-meeting remarks become law
• Leading Small Groups concentrate power on Xi (15:07) - Unlike the old Standing Committee where peers debated, leading small groups are stacked with subordinates who won’t disagree
• Ding Xuexiang is the most-trusted person in China most people have never heard of (23:12) - Heads cybersecurity office, science & tech commission, controls Xi’s daily life since 2017
• The party’s AI doctrine: brakes are as important as the engine (25:46) - Ding’s Davos speech revealed the regulatory posture
• Every internet company has a “kill switch” team (30:04) - Designated party-trusted operators who can pull the plug if AI generates subversive content
• Local government debt is 120-140% of GDP, on top of 60-70% sovereign (47:51) - Total Chinese government debt is approaching 200% of GDP
• Banking system is socialist; firms are capitalist (52:39) - Output-maximization rather than profit-maximization is encoded into capital allocation
• Selection bias on industrial policy (55:05) - Xiaomi, BYD, CATL, DJI — the most famous successes were privately funded; Huawei is the partial exception
• Rent-seeking, not GDP targets, drives over-construction (1:04:35) - James Kung’s research: selling land cheaply to princelings statistically increases promotion odds
• California vs. China: opposite political equilibria around construction (1:05:43) - In China the local secretary can cut through red tape; in California, regulations create rent-seekers who slow projects down (Robert Caro/Robert Moses parallel)
• Why financial repression liberalization wouldn’t help most Chinese people (1:10:28) - Savings deposits are concentrated in 10-20% of households; the bottom half need welfare, not higher rates
• Shih’s read on Taiwan: Xi wants it but not at any cost (1:11:53) - 12 years of restraint reveals preference; Ukraine likely raised the invasion threshold further
• Mao’s Stalin-style coalition-purging is Xi’s playbook (45:03) - Use ally A to purge enemy A, then use ally B to purge ally A; Zhou Yongkang was Xi’s first purge
• No succession plan exists (1:19:46) - And the playbook of designating a successor “always ends with the successor dying horribly”
• Foreign exchange reserves are <5% of money supply (1:21:09) - Even a 10% rational reallocation by Chinese savers would exhaust them
• Shih’s 2040 forecast: China’s economy ≈ 1-1.2x US (1:25:58) - Bearish, even before AI considerations
• Princelings have inherited political acumen (7:00) - Not just genes — their parents teach them how to read and survive political environments

Mentions

Companies

• DeepSeek / Highflyer (28:29) - The hedge-fund-spawned AI company; Xi met personally with founder Liang Wenfeng; trained on Chinese government policy documents
• Huawei (30:33) - The partial exception to “private companies always beat state ones”; received significant state funding
• Alibaba (1:00:02) - Located in Zhejiang, where DeepSeek also is; tech crackdown in 2021 contributed to lack of unicorns
• WeChat / Tencent (28:29) - Adopted DeepSeek “as fast as possible” under regulatory encouragement
• BYD / Xiaomi / CATL / DJI (55:05) - China’s most famous successes; mostly privately funded at inception
• Volkswagen (1:27:20) - Reference point: the world will not switch 100% to BYD
• TikTok / Red Note (36:00) - Examples of Chinese content companies thriving despite censorship constraints
• Anthropic / OpenAI / Vercel / Perplexity / Nvidia (39:08) - In WorkOS sponsorship segment
• WorkOS (39:08) - Sponsor; enterprise SSO, audit logs, RBAC for AI companies
• Scale AI (20:46) - Sponsor; SEAL safety leaderboards, Humanities Last Exam, Enigma Eval
• Ritz-Carlton (39:08) - Coffee analogy in WorkOS sponsorship
• Sequoia Capital (49:41) - Mentioned as the kind of US investment Chinese capital is barred from accessing

Products & Technologies

• AGI / AI / leading-edge models (0:00) - The central topic; party views AGI primarily through usurpation-risk lens
• DeepSeek (the model) (36:39) - Shih uses the open-source weights for academic research; better at finding policy documents than US models
• Claude (36:39) - “Okay” for CCP analysis but worse than DeepSeek
• Baidu (38:08) - Trained more on social media; less useful for policy research
• Falun Gong content (30:13) - Specific example of subversive content that would trigger an AI shutdown
• Local government debt instruments (46:34) - 10T RMB special debt swap last year; underlying debt keeps growing
• Belt and Road v2 (32:18) - Dwarkesh’s framing for Chinese AI infrastructure exports to Gulf states
• Made in China 2025 (57:59) - The plan that drives state finance toward LiDAR, EUV, etc.; Shih thinks the state-backed leg is dysfunctional
• Paxlovid / mRNA vaccines (1:16:57) - Experts told the government Western vaccines were better; party chose domestic regardless
• Zero COVID (8:08) - The signature example of party-protective stupidity overriding technocratic input
• One-child policy (1:18:00) - Cited as another major bad call that wiped out the smart calls

People

• Xi Jinping (0:00) - The central figure throughout
• Ding Xuexiang (6:07) - Politburo Standing Committee member, head of cybersecurity & science/tech offices, gatekeeper of Xi’s information flow
• Li Keqiang (5:10) - Former Premier; actually had real economics training
• Ma Xingrui (5:23) - Xinjiang Party Secretary, Tsinghua-trained military industrialist
• Zhang Guoqing (5:23) - Politburo, military industry, top science programs
• Liang Wenfeng (28:29) - DeepSeek/Highflyer founder; met with Xi in industry leaders meeting
• He Lifeng (33:00) - Vice premier, finance lead, US trade negotiator; built Tianjin’s “New Manhattan”
• Li Qiang (31:07) - Premier; possible alternative coordinator for AI infrastructure investment
• Bo Xilai (46:00) - The “Trotsky” figure; flamboyant princeling, now imprisoned; son in Canada and on X
• Zhou Yongkang (45:03) - Former security czar Xi purged with Hu Jintao’s help — his first big move
• Hu Jintao (45:03) - Xi’s predecessor; helped purge Zhou
• Jiang Zemin (23:19) - Earlier leader whose Shanghai stronghold Ding Xuexiang helped break apart
• Stalin (13:13) - Recurring parallel; same low-key bureaucratic ascent, same coalition-purging method
• Trotsky (45:54) - The flamboyant counterexample; Bo Xilai is the Chinese analog
• Mao (45:40) - The post-Mao Long March generation succession is the model that worked once; not replicable now
• Jiang Qing (Madame Mao) (1:22:31) - Wife who served as transitional figure; purged within weeks of Mao’s death
• Putin (1:13:41) - Whose Ukraine miscalculation likely raised Xi’s threshold for Taiwan action
• James Kung (1:04:35) - HK economist whose research links princeling-discount land deals to promotion odds
• Robert Moses / Robert Caro (1:06:15) - The American parallel for the political economy of construction
• Art Kroeber (1:01:54) - Author whose book Dwarkesh has reviewed re: Chinese infrastructure productivity

Surprising Quotes

“The party’s very, very paranoid that some actor outside, but even inside the party is going to use it as a tool to usurp the power of the party.” — Victor Shih, 0:00

“Some people die, you know, whatever — we have slower growth for a couple of years — but we get to preserve this power. We will go ahead and preserve power.” — Victor Shih, paraphrasing the party’s revealed preference, 9:00

“If you’re a startup at Tsinghua University, you have access to private sector funds, but also some of the central government seeder funds, if it falls in the right category.” — Victor Shih, 49:10

“Anyone who’s been a mid-level or higher official can no longer leave the People’s Republic of China.” — Victor Shih, 1:28:06

“I think this transition is going to be more ruthless, more brutal, and potentially more disruptive.” — Victor Shih, on what happens after Xi, 0:39

Transcript

Victor Shih: 0:00 AI apparently is one of the higher priorities for Xi Jinping. The party’s very, very paranoid that some actor outside, but even inside the party is going to use it as a tool to usurp the power of the party. The Chinese leadership, they’re not going to play golf three days out of a week. Xi Jinping and his colleagues are having meetings about policies 270 days out of the year. If Xi Jinping expresses a preference over a policy direction, that is a direction you have to go toward, no matter what. If he observes that you’re dragging your feet, you will be purged.

Dwarkesh Patel: 0:38 What is the succession plan after Xi is gone?

Victor Shih: 0:39 There are no plans right now. I think this transition is going to be more ruthless, more brutal, and potentially more disruptive.

Dwarkesh Patel: 0:49 Today I’m talking with Victor Shih, who is a director of the 21st century China center at UC San Diego. China is obviously the most important economic and geopolitical issue of our time, and doubly so if you believe what I believe about AI. I was especially keen to talk to you because you have deep expertise not only in Chinese elite politics but also its economic system, its fiscal banking policies. We’ll get into all of that. But before we get into the AI topics, is China actually a more decentralized system than America? If you look at government spending by provincial governments versus the national government in China, it’s like 85% local/provincial, 15% national. In the US, it’s actually, if you look at state and local, it’s 50% federal government and 50% state and local. When you think of an authoritarian system, you often think of it as very top-down, the center controls everything. But if you look at these numbers, it just seems like it’s quite decentralized, or is that the wrong way to look at the numbers?

Victor Shih: 1:46 I think for a while China was quite decentralized. So this is kind of in from the mid-1970s all the way until the mid-1990s, China was very decentralized where local governments generated a lot of revenue, but they also spent a lot of money. And it incentivized them to do, you know, kind of basically good things, such as trying to attract as much FDI as possible, trying to attract as much local investment as possible, give tax breaks and so on and so forth. So I think, you know, China had a very good period of mainly sort of private sector driven growth because of this fiscal decentralization. And that’s understanding of I think most economists. But then there was a tax centralization in 1994, basically where the central government basically said, okay, this is ridiculous, we don’t want to end up like the Soviet Union, you know, and fall into different pieces, we need to control fiscal income. So then they grabbed the most lucrative source of taxation at that time, and it continues to be the case, which is the value-added tax. And so the collection of value-added taxes and then also in subsequent decades pretty much all other tax categories. are now collected by the central government. Then they reimburse part of the money to the provinces. And then supposedly they say, okay, now you can spend the money as you wish, but in reality, that’s not the case. So in reality, it’s kind of like, well, if you do this thing, then I’ll give you a little bit of money, you know, but you have to do the thing that I want you to do to get this money, to get this grant, basically. Uh, so then the fiscal autonomy, I would say, has been falling since 1994. I mean, there were different periods… sometimes, you know, there’s a little bit more autonomy. Uh, so, you know, sort of from 2000 to, uh, until recently, until 2020, the localities gained more autonomy vis-à-vis land sales. So that’s when the real estate market was going very well. Even though they couldn’t control the tax revenue, they could control the land revenue. Uh, but then the central government basically killed the land market in 2022. Uh, and and now I think localities in China are highly dependent on the central government.

Dwarkesh Patel: 4:00 Mm-hmm. Let’s start at the very top. If you look at the members… first I want to understand the personnel. If you look at the members of the Politburo, uh, and you look at their past, a lot of them will be provincial leaders and then some local thing before that. But in many cases, they’ll have stints as you said, as they were the head of this, um, chemical engineering SOE or they have a, they studied… and Xi Jinping himself was a chem… like has a PhD in chemical engineering, right?

Victor Shih: 4:29 Something. Some kind of… well, but I… yeah, I’m not… yeah, I’m not…

Dwarkesh Patel: 4:32 Yeah, so that’s actually what I want to ask about. If you just look through the the list, a lot of them are like PhD economist, PhD industrial engineer, etc., etc. Um, I mean, some of them and you look it’s like, oh, I’ve… they have a PhD in Marxist theory or something.

Victor Shih: 4:46 Marxist-Leninist, yeah. Marxist thought.

Dwarkesh Patel: 4:48 Um, which that seems fake. But are they actual technocrats? If you look at their degrees, they seem quite impressive. Um, or is this just like fake made-up credentialism? If there’s like a PhD economist in the Politburo, am I… should I be like impressed and be like, oh, they probably really understand economics, or should I just think this is kind of a fake degree?

Victor Shih: 5:10 Some… well, so the former, uh, Premier of China Li Keqiang, uh, his undergraduate degree, I think was economics and he studied under one of the best economists in China.

Dwarkesh Patel: 5:22 Yeah.

Victor Shih: 5:23 Uh, so I think he understood economics. Uh, so in, I would say, in the Politburo today, there is this wing of the Politburo who are military industrialists, who were trained as engineers and they indeed, uh, were trained in the best programs in China. So Ma Xingrui, the Party Secretary of Xinjiang, for example, in the Politburo, uh, you know, graduated from Tsinghua University. Uh, Zhang Guoqing also, uh, who worked in a military industry for years, uh, he grad… I don’t know if it’s Tsinghua or one of the other top science programs in China. So they know-

Dwarkesh Patel: 6:00 Know a lot about science, but does that mean that they know a lot about governing? I mean, that’s, I think, a somewhat different question. But then you also have other cases, uh, like Ding Xuexiang, right? So, we, we can talk about him because he’s in charge of cybersecurity in China.

Victor Shih: 6:15 He has a technical background but in metallurgical forging, graduating at I wouldn’t say it’s not the MIT of China, maybe closer to the IIT of China, you know, just kind of like this second-tier science, uh, institute, uh, founded by one of the Soviet experts who wanted to teach China how to do metallurgy. Uh, but that, does that kind of very specialized technical knowledge make you a better leader? I think it depends on the person. You know, and, and I think what happens in the Chinese Communist Party is that as people rise up, there’s certainly you have to have political acumen. If you don’t have political acumen, you’re not going to make it. And some people on top of their technical expertise, it turns out that they have some political acumen, but it’s not something that you, you sort of know ex ante, except for the princelings, right? So the princelings, the reason why they’re special is not just because of their genes, you know, they’re children of high-level officials, is that they, they grew up honing this political acumen because their parents could teach them about it. Uh, so then some people turn out to have political acumen, they’re pretty smart, they don’t necessarily have the best degree, for, you know, degrees from the best universities in China, and they end up being very successful. So it’s not necessarily like, oh, you know, all the Peking and Tsinghua University graduates, you know, they all get to the very top. There’s certainly disproportionate amount of people like that in the Politburo, but they don’t dominate because at the end of the day, it is somewhat random.

Dwarkesh Patel: 7:54 Yeah. I still feel confused about how to model the competence of the party, because on the one hand, they are super educated on, in STEM and are, uh, selected at least partly on merit, etc. On the other hand, we just see them make, like, a lot of Zero-COVID was just genuinely stupid, like scrubbing down airport runways and so forth, and it just lasted for too long, and why, why was nobody steering the ship then? So…

Victor Shih: 8:30 Well no, what’s, what’s a way to put, put together this picture? I mean, so there’s a lot of expertise at, sort of, the lower level that can feed to the higher level. And there’s, there’s some channels for doing that. I would say a lot of channels for doing that. Uh, so they’re not so dysfunctional on that score, but what time and again leads to suboptimal policy is the party’s, uh, you know, sort of instinct to preserve itself. And, but, but then in a way worse than that to preserve… Preserve its power. Right so it’s not it’s not like oh well you know if we do this kind of thing, uh you lead to a better sort of outcome, public health outcome and we get to survive you know politically, but if they see it’s like okay even if we do survive politically but it leads to a diminution of our power in one respect you know we have less control over banks or like the scientists become more independent or something like that then they would say let’s not do that. You know because it’s like what’s the cost you know some people die you know whatever we have slower growth for a couple of years but we get to preserve this power uh we will go ahead and preserve power.

Dwarkesh Patel: 9:30 But if you’re an economist, okay so here’s a question. Like the traditionally the way I understand if you study economics in at least an American university you’ll learn about like supply demand here’s why certain regulations can be bad here’s why tariffs are bad. Just the basic 101 teaches you that. So if a lot of the members at the highest levels of the Communist Party had this basic understanding, then what’s the reason that they make mistakes that many economists say they shouldn’t do? Like why obviously people talk about rebalancing a lot, why don’t they think that increasing consumption is a good idea?

Victor Shih: 10:21 Well so there’s a traditional reason and then there’s today’s reason which is on top of the traditional reason. The traditional reason is that uh you know so education in China you get tracked into science STEM or you get into social science or humanities. If you get tracked into STEM you may never learn anything about supply and demand. It’s just not required. You just learn a bunch of math, you learn a lot of engineering if you want to be an engineer, uh and then there’s some art classes for you to select you can select the art classes or you know maybe one econ class. For a while you know learning economics was popular so in the 1990s all the engineering students would have taken an econ class but people who were trained before that they were not. The one class that is still required for everybody is uh government ideology. Right so um in fact the requirement for government ideology has been ratcheted up in recent years so that now because I I look at all the applications of students from China and I look at what classes they’ve taken. Now once a year they have to take a class called uh situation and policy which is basically just you know the Chinese government’s perspective on all kinds of different issues. Uh but that’s not economics. It’s not accounting. It’s not you know some basic skill it’s just telling people what the government prefers in a given topic. Uh so that’s the traditional reason so you you can have some very high level officials even if they have PhD in nuclear engineering they may not know very much about how the market works how society works so on and so forth. Yeah. The new reason is that uh we have one very very powerful figure in Chinese Communist Party such that uh even if you’re a Politburo member …which used to give you some autonomy in the domain that you govern over, that is no longer the case. You know, if Xi Jinping expresses a preference over a policy direction, that is a direction you have to go toward no matter what. Because if he observes that you’re dragging your feet or that you’re pursuing a slightly different agenda, you will be purged. And we’ve seen cases of that.

Dwarkesh Patel: 12:27 At some point we’ll talk about leading small groups and the way they work. But the picture you get from the fact that Xi is monitoring what all these different Politburo members are doing, he’s noticing whether they’re dragging their feet. He’s leading all these small teams—you’ll explain this in a second how these work—but they organize different leaders who are at the heads of the ministries or commissions that are relevant to a specific project working. And I guess every week they meet and they hammer out who is the bottleneck, what do we need to do to get a project moving. So Xi has ratcheted up the amount of such meetings and he’s… the picture you get is somebody who is micromanaging details across all these policy areas. If you read Stalin biographies, there’s a big thing like Stalin actually was a person like this. Obviously not to great effect, but he was interested in micromanaging every single detail across even the theater productions in the country to like steel production, everything. Is that who Xi is?

Victor Shih: 13:13 Yeah, so if you look at what he does day-to-day, he spends a lot of time in meetings. And it’s astonishing. I mean, there’s one thing, it’s like the Chinese leadership, they’re not going to play golf, you know, three days out of a week. I mean, Xi Jinping and his colleagues are having meetings about policies almost every day, you know, like 270 days out of the year.

Dwarkesh Patel: 13:31 So, the study sessions or the Politburo, these meetings, etc.

Victor Shih: 13:36 Yeah, yeah, yeah.

Dwarkesh Patel: 13:38 Are they like… are they substantive? Are they real? Is he like actually learning the intricacies of like, ‘here’s how zero-COVID has impacted this province and here’s why we should whatever, whatever,’ or are they just faux? Because if you read the announcements from the party departments, they’re just so vapid, like typical communist speak. How real are these meetings?

Victor Shih: 14:00 Um, so I’ve talked to some people who’ve lectured, you know, to the Politburo. I think they are somewhat real because they don’t know ex ante what questions will be asked. Like I’m talking about the speakers, right? So I think the, you know, Xi Jinping and other leaders, they know what the speaker would say in the main remarks. But then there’s a Q&A portion of it which actually is real, like it’s not staged. Uh, so that’s pretty scary, you know, if you’re some professor, you know, Xi Jinping asks you to come in and asks you a bunch of random questions. Uh, it is very scary. Um, so I think they are substantive. And then, well, but the thing is also… Xi Jinping will make remarks after such a lecture, and those remarks become policy.

Dwarkesh Patel: 15:06 Yeah.

Victor Shih: 15:07 Um, so you asked about the leading small group. So this is where this ranking’s very important, right? So the head of a leading small group always has to be the senior most ranked person. And of course, the—the good thing about having Xi Jinping as the head of a leading small group is that his authority’s not going to be challenged by anyone else because everyone agrees he’s the most powerful and most senior ranked person in all of China. Uh, but the problem is all the other members of the leading small group are lower ranked than him. So then—so whereas before, well not all, but like whereas before when decisions were made in the Politburo Standing Committee, notionally the party and also even the bureaucratic ranks of all the Politburo Standing Committee members are equivalent to each other, so there was debate, uh, there were debates and then, um, you know, even sometimes overturning of the position of the party secretary general sort of historically you have seen cases like that. But when most of the decision was pushed into all these different leading small group, usually there’s one other person who’s similarly ranked as Xi Jinping, you know, the premier of China or one of the other people, but everyone else would be these vice-premiers, ministers, or provincial governors who are definitely ranked lower than Xi Jinping. And, you know, in no scenario will they debate a particular policy with Xi Jinping. So at most maybe there’s one person with sort of equal standing officially with Xi Jinping who could voice some kind of disagreement. But of course because Xi is so powerful informally, no one would actually dare to do that. And so basically the whole thing once all the decisions got pushed into leading small groups, basically became Xi Jinping. And it is very hard on him because basically before all these meetings there’s a briefing, you know, someone gets out a briefing book and it says, ‘Okay, this is what we’re going to talk about, here are the policy options, which one do you want?’ He has to make up his mind. I mean, sometimes I’m sure he consults with other people and so and so forth. So then he has to be really engaged with a large number of policy area on a day-to-day basis because the decisions that he make becomes the law, basically.

Dwarkesh Patel: 17:30 Yeah. And it’s stupendous to think about. We understand—and you can like tell somebody how big is China, the population, etc, the size of the economy. I think visiting China gives you a tangible sense of just how big. You have provinces each of which have as many people as a normal nation state. Provinces which have hundreds of millions of people, 40,000 people, 40—40 million people, 50, you know, so forth. Um, and then one person will say something at the end of a meeting that determines the policy for all these people. It’s just an incredible concentration of power. For basically like everyone. For like the population of like North and South America or something. Does he say things at the end of meetings, during his speeches, etc., which indicate that he actually has a good understanding of these different policy areas? Because obviously the official speeches are very vague and it’s always something like we will make sure that we are robustly pursuing Chinese growth while you know hewing to the tenets of Mao’s thought and while that—so never anything specific or interesting or—but do we have an indication that behind closed doors he’s—

Victor Shih: 18:37 Well, you don’t think it’s anything specific or interesting, but if you have read, you know, thousands of these things, sometimes it’s quite revealing actually.

Dwarkesh Patel: 18:51 So I am interested. What does it tell—

Victor Shih: 18:53 Yeah, it is general. Yeah, it is true.

Dwarkesh Patel: 18:55 But I’m interested in what it reveals about him personally. Does he display any analytical ability, any sort of like empirical understanding of different policies, etc.?

Victor Shih: 19:12 Yeah, that is difficult. So I would say because of his upbringing as a princeling, for internal party matters, how to control the party, how to control the military, from the speeches, you know, some of these—they’re not secret speeches, but they’re like sort of more geared toward other members of the Chinese Communist Party. So some of these speeches, like if you go to China, you can read some of these speeches, which I have, he has a really good political nose. Like he knows what it takes to control the party apparatus, control the military. On especially economic issues, yeah, it seems like, you know, some advisor gives him some talking points, he talks through it, there is some sense of that. On technology issues, I think he cares a lot about it, but only from the perspective that competition with the United States is a very important objective of the Chinese Communist Party, is for him personally a very important objective, so he wants to win. But precisely the extent to which he understands all the intricacies, it’s unclear to me. I don’t think he necessarily—you know, I don’t think he does. But then, you know, compare—comparing him to American leaders, especially today, I think he’s probably better prepared, just because in China, the experts, they talk to the top leadership all the time.

Dwarkesh Patel: 20:38 Yeah. The advantage of the American system, of course, is that at least in theory, it can survive a leader who is not up to the snuff. Publicly available data is running out. So major AI labs like Meta, Google DeepMind, and OpenAI all partner with Scale to push the boundaries of what’s possible. Through Scale’s Data Foundry, major labs get access to high-quality data to fuel programs… post training, including advanced reasoning capabilities. Scale’s research team SEAL is creating the foundations for integrating advanced AI into society through practical AI safety frameworks and public leaderboards around safety and alignment. Their latest leaderboards include Humanities Last Exam, Enigma Eval, Multi-Challenge, and Vista, which test a range of capabilities from expert level reasoning to multimodal puzzle solving to performance on multi-turn conversations. Scale also just released Scale Evaluation, which helps diagnose model limitations. Leading frontier model developers rely on Scale Evaluation to improve the reasoning capabilities of their best models. If you’re an AI researcher or engineer and you want to learn more about how Scale’s data foundry and research lab can help you go beyond the current frontier of capabilities, go to scale.com/dwarkesh. Okay, let’s go back to Ding Xuexiang, how do you pronounce his name?

Victor Shih: 21:58 Ding Xuexiang.

Dwarkesh Patel: 21:59 Okay, sorry about that. So Politburo has what, 25 members? 20, 25?

Victor Shih: 22:05 Yeah, so 24 members, yeah.

Dwarkesh Patel: 22:07 Okay. Standing Committee has seven members. This is the the key group inside the Politburo. And this person is one of those seven. He runs the Center for Science, sorry, the Central Science and Technology Commission. For my, the reason I’m especially interested in him is because any large-scale AI effort that China would launch would be under his purview. Do we understand what he wants, who he is, what his relationship with Xi is, what he would do if AGI is around the corner, etc?

Victor Shih: 22:41 Yeah, these are all very good questions. Um, so I think in addition to this Science and Technology Commission that you mentioned, another very important position that he holds is that he’s the head of the office of the Central Commission on Cybersecurity.

Dwarkesh Patel: 22:59 Hmm.

Victor Shih: 23:00 The head of the Central Commission on Cybersecurity is Xi Jinping himself, but the head of the administrative office, which runs the day-to-day operation of that commission, which basically a leading group, is Ding Xuexiang.

Dwarkesh Patel: 23:11 Yeah.

Victor Shih: 23:12 So, and he took that position, uh, back in 2022, so he’s been steeped in cybersecurity for a couple of years.

Dwarkesh Patel: 23:18 Yeah.

Victor Shih: 23:19 I think by now he knows all the major players, uh, and some of the key policy issues. Uh, his relationship with Xi Jinping is very interesting one. So he only directly worked under Xi Jinping for one year, when Xi was in Shanghai. He was basically the very senior level secretary who would support whoever the Party Secretary of Shanghai was. And he worked under three different Party Secretaries of Shanghai, uh, supported Xi Jinping, uh, for one year. For some reason, Xi Jinping just trusted him absolutely after that. is a mystery and you know I’ve looked at all the open literature, I’ve asked people in Shanghai, nobody knows why that’s the case. I mean, almost certainly I think one reason is that he was gathering a lot of information about the other leaders of Shanghai and sending all that information to Xi Jinping to, uh, basically let him know because at that time, as you know, the previous leader of China, Jiang Zemin, had a very big stronghold in Shanghai, so basically Xi Jinping had to sort of break that apart before he could assert control over Shanghai. Ding Xuexiang likely was one of the people who sent all this necessary information to Xi Jinping in order for him to do that. But beyond that, I mean, lots of people do that for Xi Jinping, truth be told. So beyond that, what else has he done for the big boss to earn his trust? I think is a big mystery. Uh, but the true, you know, but the manifested outcome is that Xi trusts him a great deal because in 2013, he was promoted to Beijing to be the second in command in his personal office, handling all the flows of data in and out of the desk of Xi Jinping. So that’s a very important position. He became the first in command couple years later. He took control over the entire apparatus that governed Xi Jinping’s day-to-day life back in 2017. Uh, now he’s a vice premier in charge of cybersecurity. So clearly this guy is trusted.

Dwarkesh Patel: 25:31 Hmm.

Victor Shih: 25:32 Uh, you know, his preference for AI, AGI. So, uh, I, I tweeted this, uh, Davos speech which you, you also looked at. Um, I think it’s very revealing, right? So you look at you’re like, ‘What’s…’. So what he said was, ‘Uh, we need to invest in AI, but we need to do it, we can’t go all out in investing in it without knowing what the brakes are. We have to develop the brakes also at the same time.’ I think that’s extremely revealing. So basically, it’s very different from the American approach because in America, it’s all driven by the private sector and of course all these, you know, except for one or two companies, everyone’s just invest, invest, invest and try to reach AGI as soon as possible. For the Chinese government, they’re very afraid that some actor outside, uh, but even inside the party is going to use it as a tool to usurp the power of the party. So they want to know that they have a way of stopping everything if it comes to it. So for them, developing the brakes is just as important as developing AI itself.

Dwarkesh Patel: 26:40 Hmm. What are the different organizational milestones we should be watching for to understand how they’re thinking about AI, when things have escalated? So to add more color to this question, what I expect to happen, let’s say next year, in terms of raw AI capabilities, we might have computer use agents. So a thing which is not just a chatbot, but can actually do real work for you, like… put it loose on your computer it’ll like go through your email and compile your taxes etc. and then dot dot dot in five years I expect it to be able to do all white collar work or most white collar work, which is like you know 40% of the economy potentially. eventually we’ll have full AGI even robotics and so forth. As this is happening how should we be tracking how the party is thinking about what is happening how seriously they take it what they should do about it should we I think last week or a couple weeks ago there was a Politburo study session on AI but I don’t know if it was like AGI AI or what kind of AI they were discussing

Victor Shih: 27:34 Mostly AI.

Dwarkesh Patel: 27:35 Should we be looking for a leading group on AGI in particular should we be trying to just read more speeches what should we be paying attention to

Victor Shih: 27:44 Uh so far it seems like it’s under cybersecurity leading group. So Ding Xuexiang would be in charge of it. There won’t be a new group. I don’t think so. because for the party it’s the security aspect of it that’s the most important. I mean there’s one aspect where it’s like yeah it’ll be great to automate all industrial production. China’s doing a good job doing that already. I don’t think they need additional infrastructure you know sort of governing infrastructure to allow them to do that. but on sort of applying you know AGI or AI to governance to even service sector you know generating video content for people doing you know even like travel agency stuff the party is very very paranoid that some hostile actor outside of China certainly they believe that very strongly or inside of China is going to do something and the AGI’s going to take off and it’s going to undermine the party’s authority. so the what what they I think basically what we’re going to see in terms of institutional development is not at the top end but on the lower end. so they will want to designate human beings in all the government agencies in all the commercial entities that are using AI or AGI to put their foot on the brake if it comes to it.

Dwarkesh Patel: 28:29 But so far after the Deepseek became a big deal the immediate reaction it seems in China was the all the different big tech companies WeChat Tencent everybody was encouraged to adopt it and they did as you know adopted as fast as possible Xi himself met with Liang Wenfeng and did a whole meeting where the all the industrial heads were there and Xi said we have to accelerate I don’t know whatever technology in China etc. so it seems like so far the response has been the greater the capabilities of AI the more they’re excited about it the more they think this is the beginning of Chinese greatness but you think that changes at some point?

Victor Shih: 29:44 No no I mean they want to develop it of course they’ll pour a lot of money they’ll try to give Deepseek as much help as possible you know sourcing GPUs you know whatever all this kind of stuff but I am all but certain that there is one and maybe a team of people

Dwarkesh Patel: 30:00 In the headquarters of DeepSeek, who can pull the plug?

Victor Shih: 30:02 Interesting.

Dwarkesh Patel: 30:03 If necessary.

Victor Shih: 30:04 Because there’s such a team of people in every major internet company in China.

Dwarkesh Patel: 30:09 Yeah. What would cause them to pull the plug? What’s an example of…

Victor Shih: 30:13 Uh, if subversive, yeah, so if AGI started generating, you know, Falun Gong related content and it proliferates very quickly beyond the capacity of the censor to control it, then they’ll have to stop, you know, the, you know, the algorithm from generating new images and and new videos and so on and so forth.

Dwarkesh Patel: 30:33 Hmm. Suppose it turns out you were saying look, they can help high-flyers source GPUs and etc. What is the mechanism by which, suppose it turns out that for Highflyer to continue making—DeepSeek to continue making progress, they need all the GPUs that Huawei can produce and they need a whole bunch of other things, they need energy, they need data centers. What would it actually look like for the government to say every single Huawei GPU has to go to DeepSeek, they’re going to get access to all this spare land to build data centers? Who would coordinate that? Would that be the leading cybersecurity group? Would that be somebody else?

Victor Shih: 31:07 Yeah, it could be, yeah, so that’s an interesting. So the, um, the physical investment will have to share power with some other leading group. Yeah, so I guess that would be a justification for an AGI leading group of some sort because, you know, building power center that falls under NDRC, it could be under Li Qiang or He Lifeng, um, you know, which especially He Lifeng who is in charge of like investment and financing, uh, so but if, if Xi Jinping doesn’t want He Lifeng to share that power then we’ll have to give it all to Ding Xuexiang and then give him a lot of power. I mean he can just do it by fiat or create a separate organization to make it happen. Uh, at this point as you know, you know, data centers are being built in China and a very fast clip and I think they’re just using existing command structure. Uh, there is potentially a foreign aspect of this, right? So it’s like, uh, you build up all this cloud computing capacity in the Gulf states, who’s going to be the customer? You know, who’s willing to pay billions of dollars to to use it? China, right? So, uh, so then then the foreign policy apparatus will have to coordinate that and so on and so forth.

Dwarkesh Patel: 32:18 Belt and Road V2.

Victor Shih: 32:22 Exactly.

Dwarkesh Patel: 32:26 Does it have any end implication to the progression of AI who actually ends up controlling the these um these leading groups, whose purview it ends up under? Or is it just basically a detail but in terms of, you know, what happens with AI in China it doesn’t have that big an implication?

Victor Shih: 32:45 I think it it does matter. Yeah, so you have some decision makers who have shown to be not very good, uh, and even politically not very good, uh, but nonetheless for some reason they’re trusted. I mean precisely maybe precisely because they’re not very good, they’re very… …dependent on the on Xi Jinping that they get promoted. The chief negotiator with the United States, He Lifeng, vice premier in charge of finance. He is well-known for starting and perpetuating the largest real estate bubble the world has ever seen in Tianjin. I don’t know if you went to Tianjin, there are like just empty buildings everywhere. So there’s the ‘New Manhattan’, you know, literally the size of Manhattan, filled with empty office buildings.

Dwarkesh Patel: 33:30 It ironically sounds like a great tourist site, not because it’s a real Manhattan, but because it’s so empty.

Victor Shih: 33:34 Yeah, yeah, exactly. It’s actually cool to visit, but you know, it’s like, ‘Wow, this is great, like what are you gonna do with the building?’, you know. We went to… this was in Anshun, we went to this huge, huge Buddhist temple. By huge, I just mean like, there would be this structure, the shrine, and you’d go through it and you’d realize behind it was an even bigger shrine that was hiding from view. And then you’d go through that one, there’d be an even bigger one. This happened like five times. If we were to drive through this, it would take you like a while.

Dwarkesh Patel: 34:05 But it’s new, right? Not historical, right?

Victor Shih: 34:08 Oh, no, yeah. It’s new. And then I asked like the head monk… by the way, there was nobody else there. It was me, three white guys, and nobody else. And I asked the head monk, ‘How did you guys finance this?’ And he says, ‘We’ve got a lot of supporters. A lot of donations.’

Dwarkesh Patel: 34:25 I don’t think so, yeah.

Victor Shih: 34:28 Yeah, so, so then there it really depends, like who’s in charge of AI. Yeah, so Ding Xuexiang really is a mystery. I think he has great political acumen. He must, you know, in order to gain Xi Jinping’s trust. He has some technical background, but it’s metallurgical forging. But he’s worked in Shanghai for decades, so he knows how private corporations run, he knows a lot about international trade, FDI. So I think, you know, if you look at the sort of people in the Politburo, Politburo Standing Committee, I would say he’s definitely in the top quartile of people that you would want to be in charge of AI, if not the top, you know, two or three people.

Dwarkesh Patel: 35:13 Right. Interesting.

Victor Shih: 35:14 I mean, one thing that I would say about AI is that content creation is a big, is going to be a big roadblock for China because China is so paranoid about the content that flows on the internet that they put human beings all over the place to slow things down, slow the transmission down. And now if you have AI creating, you know, tons of content, there will still be human beings, algorithms and human beings double, triple-checking the content because they’re just so afraid of some subversive content getting out.

Dwarkesh Patel: 35:59 But it’s still been compatible with them innovating and having leading… companies in even content right like Red Note and TikTok and so forth not to mention WeChat so if AI ends up and AI might be even more robust to

Victor Shih: 36:13 some of these yeah

Dwarkesh Patel: 36:15 sensitive content because you can just teach the AI to not say certain kinds of things especially

Victor Shih: 36:19 but even if that were the case even if you train something that you think is pretty good they will still want a human being checking everything.

Dwarkesh Patel: 36:25 as somebody who is trying to observe what the party is doing here have you found that using LLMs has been helpful and if so which model gives you the best insight into what the CCP is thinking?

Victor Shih: 36:39 Yeah no I’ve been using AI quite a bit I mean both to sort of code data you know used to be like if text hand coded stuff and now a lot of it can be automated but in terms of finding out what the Chinese government is doing I think some of the American models are okay like Claude is okay but the most helpful has been DeepSeek so to me it’s quite clear that DeepSeek which of course was first developed by a hedge fund to trade in the Chinese market part of what they were they really trained the model for was to detect important policy documents and meetings within the Chinese government’s because when you enter you know the right or even not even like especially sophisticated prompts you know just like you know what is the Chinese government doing when it comes to AI it comes out with a bunch of very high quality links which is very useful for my research so then you know immediately you get a sense of like what are the latest policies what are the latest statements by high level officials it seems that DeepSeek puts some you know higher weighting let’s say on on this kind of content so so I have you know I don’t dare to install it on my phone but you know I certainly have collaborators who have downloaded the open source model and install into a hard drive and I use the web interface for some of my research.

Dwarkesh Patel: 38:02 Do you think it’s just because they have more Chinese language data and so it’s just better at understanding the party communications?

Victor Shih: 38:08 So I’ve used um you know sort of Baidu’s thing and so I think the typical Chinese language models are more trained on social media content which will you know come back with hits that are like more social media’s like oh you know if you ask about AI will talk about like how AI gets used in different things instead of this kind of policy documents and meetings and so on and so forth so it it seems you know just from my inductive observation that DeepSeek you know is trained on this kind of content.

Dwarkesh Patel: 38:37 Mm so that Highflyer can make trades.

Victor Shih: 38:40 Yes so that Highflyer can I mean you know obviously China being China government policies have a huge impact on how different stocks do and that’s often where you find alpha in the old days people used to call up their friends who worked for this and that ministry to get insider information but I think what Highflyer has discovered would also generate alpha is to really use algorithm to look closely at these policy documents.

Dwarkesh Patel: 39:08 Look, coffee isn’t the reason that you’re going to stay at the Ritz, but it could be the reason that you leave. Because when you offer truly premium product, your customers require excellence in all the parts, not just the core. So, the Ritz works with a specialty roaster to ensure that they can deliver great coffee. That’s what WorkOS does for your product, but for enterprise features, like single sign-on, access controls, and audit logs. These features are not the reason that people will love your product. But if you don’t have them at all, you simply won’t get a seat at the table with your biggest, most important customers. Companies like OpenAI, Anthropic, Vercel, and Perplexity all rely on WorkOS to serve their enterprise customers. In fact, Nvidia might be the only other vendor that all of these AI companies share. And that’s because not only are these features critical, they require excellence. WorkOS makes it super easy to plug in these features into your product and get back to delivering the premium experience that your customers expect. If you want to learn more, go to workos.com and tell them that I sent you. All right, back to Victor. Okay, and then this is going to sound like a naive question, but we’re talking about how the party in the abstract wants to maintain control over every aspect of Chinese life and Chinese economy. But of course the party is made up of individuals. And many of these individuals, Xi himself, went through periods where the party had extraordinary control over people’s life during the Cultural Revolution and so forth. And they personally and their families suffered as a result of this. And they’re educated people, many of them have been in industry, right? These are not naive people. So, maybe I’m just projecting too much of my Western bias here. But what is the reason that they think it’s so crucial that, you know, the banks are run by the party and the AI is run by the party and nothing escapes the party state?

Victor Shih: 41:05 Um, yeah, so the Cultural Revolution generation, I think, you know, a lot of people suffered horrendously, including Xi Jinping himself. Uh, but then basically you have two types of lessons that people drew from it. So the first lesson is like, oh, you know, the party’s too dictatorial, you know, China needs to liberalize and so on and so forth. So you have a lot of people from that generation who feel that way. Uh, many of them now live in the United States, you know, basically they, they tried to leave China as soon as possible in the 1980s and succeeded. Uh, but for someone, apparently for Xi Jinping himself and others like him, the lesson was just don’t be on the losing side. You know, in any political struggle, make sure you’re on the winning side, because if you’re on the winning side, then you can do terrible things to your enemies. Uh, so that apparently is the lesson that he learned. And he basically honed his skills and built his coalitions for, uh, the time that he would take over high-level positions in the party. party for decades. As it turns out, so if you look at the data, you know, he uh when he was in Fujian Province he spent an inordinate amount of time hanging out with military officers in Fujian back in the 1980s, you know, late 1980s, early 1990s and then you’re like well why did he bother to do that? You know most local officials they don’t do that. I mean you know the military’s there you need to be nice to them and all this stuff but like he built a dormitory for them, he even joined a unit of uh anti-aircraft regiment, he tried as much as possible to do as much as possible with the military because he knew he needed the support of somebody in the military when the time comes and guess what uh sort of 30 years later when he was about to be elevated to be Secretary General of the Chinese Communist Party, many of the people that he knew back in Fujian are now generals, they’re in charge of important military units and once he came to power then he really promoted some of these people. Uh but then lately of course he’s purging some of these people which is a really interesting.

Dwarkesh Patel: 43:13 Yeah, yeah.

Victor Shih: 43:14 Yeah, he had a strategic vision about his own career trajectory and he pursued that uh in a very determined fashion I would say.

Dwarkesh Patel: 43:17 What is the main difference between him and Stalin in like sort of political maneuvering sense? Because all of this sounds very similar to…

Victor Shih: 43:24 Yeah, I would say early in their career uh they’re very similar. So uh outwardly they’re very low key because you know Stalin was the bureaucrat, very low key, not flamboyant, not making, you know, loud speeches like Trotsky did uh but you know low key getting things done uh seen as a very reliable person. Uh this was a reason why Stalin was chosen um in the first as far as I know you know I don’t know that much about Stalin. Same thing with Xi Jinping. So uh his father uh belonged to actually the liberal faction in the Chinese Communist Party. Uh his father never offended that many people within the party. I mean you know everyone at a high level will have you know sort of screwed over somebody along the way but his father was in in the better category let’s say. Uh and he himself was very low key while the other princelings were fighting for very good jobs in the capital city of Beijing or in Shanghai, Xi Jinping said oh no it’s okay I’ll go down to the villages, I’ll work in Hebei in a rural area then Fujian which was kind of a peripheral area uh and then Zhejiang was slightly better but still not Beijing or Shanghai. So he got out of the way of this political infighting. Uh I think similar in a similar way that Stalin did like he early on he did not confront people too much. Uh but once he came to power he knew they both knew what it took to control the the party which they they govern over. Uh basically you form a whole series of coalitions to get rid of your most… threatening enemy at a given time. Right? So for Xi Jinping it was like, oh, well, you know, this guy Zhou Yongkang who controlled the police forces and also the Ministry of State Security at the time, doing all these irregular things, being fabulously corrupt. He’s a threat to the whole party. So he convinced Hu Jintao to join with him to purge Zhou Yongkang, which was successful. Uh, you know, Stalin did the same thing, you know, to Trotsky. And then after the most threatening person is gone, you form another coalition to purge the next person and so on and so forth. So they both basically did that until they achieved absolute power within the party.

Dwarkesh Patel: 45:40 Yeah, yeah. And if you read the Stalin biography, so many of the, in the early part, like half the people in the Politburo are later going to end up in the Gulags and then many of the people in the party later have been to the Gulags. I wonder if there’s a Trotsky analogous person in this story, somebody who was a flamboyant speechmaker?

Victor Shih: 46:00 Bo Xilai.

Dwarkesh Patel: 46:01 Bo Xilai. Okay.

Victor Shih: 46:02 Yeah. Yeah. So he was a one competitor with Xi Jinping. But because he was so high profile that he never had a chance to be the top leader of China. Uh, and then of course Xi Jinping made sure that he would fall.

Dwarkesh Patel: 46:14 Right. And now he’s in prison.

Victor Shih: 46:15 And now he’s in prison. Yeah, he’s alive. His son is in Canada actually. He’s on Twitter now, on X.

Dwarkesh Patel: 46:19 I should reach out honestly. That’d be an interesting person to talk to.

Victor Shih: 46:21 Yeah, yeah. I think that would be really interesting actually.

Dwarkesh Patel: 46:24 Let’s move on to another topic I know you’ve studied deeply, which is the local government debt situation.

Victor Shih: 46:30 Sure.

Dwarkesh Patel: 46:31 What is the newest or what is your sense of the situation now?

Victor Shih: 46:34 Yeah, it just keeps growing in absolute size. Uh, because basically, uh, China is trying to tell the world that they don’t have a high debt level, which at the central level that is sort of the case, you know, it’s still 60, 70% of GDP. But the reason why that is the case is because, uh, there are a lot of things the Chinese government would like to do, they push it down to the local level. But the authorization of local debt issuance is authorized by the central government. So at the end of the day, it’s really the central government telling local governments like, okay, you need to do this thing. I’m not going to give you money, but I will give you the authorization to issue even more debt. Uh, so the debt level keeps going higher and higher. The, the one thing that has alleviated the cash flow pressure, uh, late last year was the central government, uh, authorized again the issuance of close to 10 trillion RMB in special local debt to repay some of the higher interest-bearing debt of local governments. Uh, but that’s an accounting exercise. It doesn’t literally decrease local debt. It just changes from higher yielding to lower yielding. Meanwhile, the overall size of local government debt in China, I would estimate to be 120% of GDP to 140% of GDP.

Dwarkesh Patel: 47:51 Whoa. That’s atop the 60% that is owed by the central government.

Victor Shih: 47:56 The central government. Yeah. So total governmental debt is, yeah, it’s, you know, pushing 200.

Dwarkesh Patel: 47:59 And the high level way to understand what this debt was for. The debt in Western countries and other OECD countries is often because you owe money basically to pensioners. In this case, it’s — um, it’s — it was just that they build bridges.

Victor Shih: 48:12 Borrowed. Yeah, they built all the high-speed rail, all the infrastructure, the beautiful infrastructure that you do see in China. And then more recently, industrial policies, right? So, for example, if there’s a new science park for AI, let’s say, right? The land that they acquire, the infrastructure that’s built is all built with borrowed money. Then there are these startups, and then the — the startups are partially financed by some of these central level investment funds that you read about, industrial policies. But then actually the local government also have their own seeder funds, which uses borrowed money to finance some of these venture deals.

Dwarkesh Patel: 48:54 For startups, is the fact that they can get cheap credit from banks more relevant? Is the fact that they can get — um loans from provincial funds more important? There’s also these central government funds, the Big Fund and so forth. Is that the — which — which is the bigger piece here?

Victor Shih: 49:10 It — so a lot of it depends on like your social network and your connections. Yeah. So if you’re a startup at Tsinghua University, you have access to private sector funds, but also some of the central government seeder funds, if it falls in the right category, you know, semiconductors, AI, etc, etc. But if you’re — at Lianwenfeng, I don’t know, so Lianwenfeng because it started out as a financial institution so it was mainly sort of private sector money from — from financial investors. But there are other AI, especially sort of more hardware-driven initiatives in the provinces which are seeded by local government funds. Uh, and — and the point about financial repression is that without financial repression, an investor in China can invest in Silicon Valley. They can go to Sequoia, invest money, but because of capital control, they cannot legally move more than $50,000 a year out of China, they have to find domestic investment. Uh, so as a result, many of them choose to invest in Chinese high-tech, or they choose to put their money in the bank. So actually a lot of people just choose to put their money in the bank, and then the state banking system is — uh then financed to — or they have the deposits with which to lend to local government seeder funds in — for industrial policies.

Dwarkesh Patel: 50:30 And the crucial thing here is that the banking sector is totally controlled by the state. So you’re not getting a true rate of return based on what the productive value of investment is. It’s literally like 1%.

Victor Shih: 50:44 Yes, the deposit rate’s 1%.

Dwarkesh Patel: 50:46 Yeah. Which — which is — I mean, if you’re — it’s like an insanely low return.

Victor Shih: 50:49 Well, inflation rate is very low in China, right?

Dwarkesh Patel: 50:51 There’s a couple mixed-up questions I want to ask here, but basically, so you said, well if there were no capital controls, these people might want to take their money and go invest in Silicon Valley. Very basic development economics would show you that like if a country’s less developed, they’ll have higher rates of return The Silicon Valley would want to invest in China. Um, why would getting rid of capital controls make capital go the other way? And this goes into other questions of why is it the case that the Chinese stock market has performed so badly even though the economy has grown a lot? And then why is it the case that um, they have a current account surplus and they’re basically accumulating T-bills when in you the pattern you should see obviously is that like they can earn much higher rates of return building productive things in China than accumulating 4% yield, uh, T-bills.

Victor Shih: 51:34 Well, so at the deepest level, I would say there’s there is a deep fundamental difference between capitalism and socialism. Uh, and it sounds very philosophical, but I think, you know, this might, uh, help sort of your Gen Z listeners, uh, think about this. Because for socialism, they only care about output. You know, it’s like, okay, uh, whatever capacity we’re thinking about, whether it’s grain production or metal metal production or these days robotics, we just want more of it, right? So then, uh, the state can use the state banking system which they control to allocate huge amount of capital to maximize the output of all these different things that they care about. Um. But when you maximize the output, you don’t necessarily make money doing that. Whereas capitalism wants to maximize profit, which is, you know, as you know, the difference between the cost of production and the amount that you can earn from selling the output. Uh, for socialism, they don’t care about that.

Dwarkesh Patel: 52:36 But the companies aren’t socialist, right? The companies are profit seeking.

Victor Shih: 52:39 Well no, but because the financial system is socialist, they’re forced into socialist-like behavior. You know, you can’t go into a bank and say, look, I make robotics, uh, this robot that I’m going to make is going to be highly profitable, um, you know, down the road but I can only make like 10 of them. And then the bank would be like, well this is BS.

Dwarkesh Patel: 52:54 I know a lot of startups don’t make any money and then at some point notionally you have to start making money. Whereas the socialist banking system…

Victor Shih: 53:00 Basically says, you know, even if you never make any money or hardly any money, that’s okay. Uh, as long as the Chinese government tells us that this is a strategic sector, as long as you can prove to us you can actually produce the thing that we want you to produce. If you never ever make money doing that, that’s perfectly fine.

Dwarkesh Patel: 53:26 What is the sort of outer loop feedback? The thing which keeps the system disciplined? Here there’s private investors who care about actually earning a rate of return and so they’re only going to invest in companies which they think have a viable shot of becoming big huge companies that can actually be at the frontier in technology. If these banks are not actually competing against real investors—

Victor Shih: 53:50 And it’s not their money.

Dwarkesh Patel: 53:51 Exactly. But it seems like this is compatible with many of the world’s leading companies being developed in this system. So— How is it working?

Victor Shih: 54:01 Well, so this is where the bureaucracy steps in. Right? So it’s like well how would a bank tell what’s a good company and what’s not a good company? Uh, so then there’s a complicated system where the ministry of, so let’s say for robotics or for semiconductors, uh there’s a sort of expert group in ministry of uh industry and informatization, information technology, that will assess uh all these different projects. So the banks will literally send them, say okay this company wants to borrow 10 million dollars, do they have real technology? Is this a good prospect? And some bureaucrat supposedly with no financial stake in the company, that’s completely neutral, supposedly, I say supposedly because in reality we’ve seen many cases of corruption, uh in conjunction with an expert panel who again with no any you know relationship with the company or with the bank, will sign off on these projects or reject these projects, get sent back to the bank and bank says okay you know these bureaucrats say that your project is good, you’re good to go, here is 100 million dollars.

Dwarkesh Patel: 54:59 But it seems to work. Yeah, yeah. No, so… but like doesn’t this just belie the whole idea that central planning isn’t supposed to work, right?

Victor Shih: 55:05 No no, so so there’s selection bias. We see and pay attention to the cases that work and there have been some fabulous success cases, which uh by the way does not include Xiaomi or BYD because those were mainly privately funded at the beginning.

Dwarkesh Patel: 55:27 And CATL and DJI.

Victor Shih: 55:28 Yeah, exactly. So the success cases actually are not that. I mean, but there are, you know, I’m sure there I I don’t know the cases but like there are some well uh Huawei, so I would say like Huawei it is there was a lot of state funding along the way and so on and so forth. So you have success cases but there are so many failed cases and billions and billions of dollars just thrown down a hole basically for the failed cases. Uh in fact for semiconductor things got so bad that they had to arrest a whole bunch, dozens of people in the ministry of information and or industry and information technology who were in charge of approving these semiconductor deals. They all went to jail because they were in a deal where you know some company submit a bogus project, they approved it, they got a part of the money and so on and so forth. Uh so it doesn’t work very well. I mean of course you have fraudulent cases in the US also you know all that kind of stuff, but but the scale in China, the waste in China is a much larger scale phenomenon.

Dwarkesh Patel: 55:54 Yeah. And it’s all financed as you say with this financial repression, which is basically a tax on savers. And so, even though there’s no meaningful income tax and just value added tax and corporate income tax, if you factor in the currency devaluation, which is a tax on consumers and then this financial repression which is a tax on savers, it actually might be a meaningful decrease in the quality of life of an average person.

Victor Shih: 56:08 Oh yeah. No, I mean, you definitely see it. Like, you know, for an economy that’s been growing… 5+ percent for decades, you by this time you would expect, you know, pretty much the vast majority of the population in China having all the basic necessities, shelters, medical care, etc., etc. You don’t see that. You know, a lot of migrant workers, they’re barely getting by. The homelessness is increasingly a problem in China, especially with the labor market being so fickle these days. Elderly care is pretty abysmal, you know, very basic level, let’s say, for a lot of people, not for everyone. So some people get very good elderly care if they worked for the government or SOE previously. So you see a lot of problems that the Chinese government has overlooked. So this is why I’m saying they have a goal of making all Chinese people very happy and fulfilled. One part of it is being fulfilled, the technology part, but the other part of truly making people living a good life, they’re not making as much progress as they should, I would say.

Dwarkesh Patel: 57:59 And then just to close the loop on that. So I think sometimes we get impressed when we see, oh, they’ve got the Made in China 2025 and the NDRC and the MIIT and so forth, have these very specific, you know, LiDAR needs to get to this point, and we need the EUV machines at this point and so forth. But your claim is that that part of the system is actually dysfunctional, and in fact the part that actually works is the 5% that is totally private.

Victor Shih: 58:31 No, so the private sector stuff works a lot better. You have occasional successes from the state finance system. So I wouldn’t say that there are no success cases, but I would say that for every success case you see, there are just maybe even over a dozen failed cases, which, you know, of course you have failed cases everywhere, you know, in VC. But the amount of money being wasted in the state financial system would be much larger.

Dwarkesh Patel: 58:59 Going back to the local government debt burden. There’s a world where AI isn’t a big deal and there’s a world where AI is a big deal. If we live in the world where AI truly causes huge uplift to economic growth and so forth, is it possible that this problem doesn’t meaningfully harm China’s ability to compete at the frontier? And here’s why that might be the case. It seems like this problem is especially affecting poor provinces, whereas you have the actual numbers, but are Shanghai and Guangdong in a good position with respect to their fiscal situation and could they keep funding, you know, top-end semiconductors, AI firms, data centers? And so China could still keep competing at the frontier in AI even if the poorer provinces don’t have the money to fund basic services and so forth. And in the long run, they get advanced AIs and that’s a big uplift to economic growth and so on. They can just grow their way out of this debt problem.

Victor Shih: 1:00:02 Yeah, so we can talk about that. I’m a bit skeptical about that for the case of China, but yes, indeed, as you said, some places like Shanghai and Guangdong, they’re still in relatively okay fiscal position. But then there are other sort of wealthier provinces with a lot of debt also. So Zhejiang Province, which is where DeepSeek is located, actually, and also Alibaba and all these tech companies, they have a high debt load, but because they start from a pretty wealthy place, they can still service the debt without too much difficulty. But actually, the trade war is going to make it worse for Zhejiang Province because a lot of the manufacturing export firms are also in Zhejiang. But nonetheless, the case of China is that basically the financial system is geared toward fulfilling the priorities of the party, i.e., the priorities of Xi Jinping. So even if they can’t do anything else, they will do the things that Xi Jinping wants them to do. And AI apparently is one of the higher priorities for Xi Jinping because he’s held multiple study sessions, he’s had different Politburo meetings and so on and so forth. And Ding Xuexiang, who’s one of his most trusted lieutenants, is in charge of it, etc. So I do expect a fair amount of resources being devoted to it regardless of the situation in local debt. I mean, in fact, at the local level things are so bad, but they just don’t care, right? So there’s civil servants, teachers who are not paid for four or five months out of the year. And you can say, well, how can you run a government like that? Wouldn’t they just collapse? But then you have to look at the outside options. You know, for your teacher, for your very low-level bureaucrat, what option do you have? I mean, do you want to go back to delivering food to people? Or would you rather at least take six, seven months of civil service salary, at least you have healthcare, at least you can get your free lunch at your workplace cafeteria every day? People still choose to work for the Chinese government, even if that were the case. But a lot of the previously very generous fringe benefits and so on and so forth, that has all been cut down because of the local debt issue, but they are still pouring a lot of resources into national defense and AI and technology.

Dwarkesh Patel: 1:01:21 And what would a solution at this point look like? Because if you’ve already spent all this money, first of all, actually, why is it the case that these investments aren’t productive? Because it’s not like it’s entitlement spending where you’re doing it for, I don’t know, old people, but then at the end of the day, you’re not getting any return for it. Theoretically, these are like, you’re building real bridges, you’re building airports. So why aren’t they productive? And then what does a solution at this point look like?

Victor Shih: 1:01:54 Well, so it was very productive, you know, I mean, I think you noted in this Art Kroeber book review that you did, in the 1980s and 1990s because China lacked a lot of infrastructure. But once you build the first high-speed rail line… rail between Beijing and Shanghai, you built a second one, you built a third one, you know, there’s very rapidly diminishing return. Also as you know, the population is basically shrinking in China, and especially in a lot of cities in northeastern China and also in southwestern China. Uh, so you really don’t need high-speed rails connecting those cities to other places in China because nobody lives there anymore.

Dwarkesh Patel: 1:03:25 Can I ask about this? So if the reason that the provincial leaders kept doing this construction was because they want to get promoted, um, and they want to get, you know, GDP numbers on the board and government spending now adds to the GDP, even if the ultimate thing you’re building is not that productive. But then the central government has had a problem with this for a while and then they think this is fake, they don’t want to keep this fake growth going, and they see the local debt numbers. Were they still promoted nonetheless? Like why, why were people doing this? Who, who was the person who created… and also, a lot of these people got arrested in the aftermath, right? So they weren’t promoted, they were actually arrested. Um, so what was incentivizing them to keep…

Victor Shih: 1:04:10 What? Uh, so when you invest in a large infrastructure project, a lot of money changes hand. And that is so so well, so this is a big debate in in my field also, you know, some people say, well look, you know, like the more you invest, more likely you’re to get promoted. My argument is that it’s not the actual investment.

Dwarkesh Patel: 1:04:34 Interesting.

Victor Shih: 1:04:35 It is the rent-seeking that comes along with the investment, right? So you know, you have to get a contractor for cement, you have to contractor for this and that, uh, you know, you get a big envelope under the table, you know, as a kickback. Then that allows you to pay your superiors, you know, because you let’s say you did a billion dollar investment to build light rail or something like that in a city, you got a hundred million dollar payout from all the contractors, you can give your superior fifty million dollars to thank that person and to up your chance of getting a promotion. And so I would say that’s more the realistic mechanism for linking investment and promotion than like, ‘Oh, good job you, you know, you generated investment and GP in this city, we’re going to reward you with a promotion.’ Uh, in fact, there’s really good work by James Kung showing this, you know, it’s like basically, and also like when you do real estate deal, you can sell land cheaply to politically connected princelings and then they will lobby on your behalf for your promotion and statistically there is an effect. So if you sell land cheaply to princelings, your chance of promotion goes up.

Dwarkesh Patel: 1:05:43 I find it fascinating the fact that you can whether a corrupt political equilibrium leads to more construction or less construction because the political equilibrium in America today is that I don’t know, California isn’t the least corrupt country state in the country. Um, but the political equilibrium is that… fact that there’s so many different factions who, you know, get their hand into the pocket means that it’s not that everybody’s like, “We need California rail to happen yesterday because I need my share of it,” it’s that, “Oh, I’m going to get a little consulting fee to slow this down by five years,” and another person… So it’s quite interesting. And if you read Robert the Caro biography of Robert Moses, and you look at the strategies he employed, it actually is very similar to what you’re mentioning here, where it’s in every single faction’s interest that a bridge that Robert Moses working on gets done through some of the mechanisms you mentioned. For example, he would give the banks these discounted bonds for this construction, so the bank lobbyists were encouraged to keep the construction going, and he would give obviously the unions wanted employment, so they’d want the construction to happen. Every single person at every stage was incentivized. So I find it interesting that China has maintained to keep a political equilibrium which encourages too much building, whereas our political equilibrium is that it’s discouraged.

Victor Shih: 1:06:54 Yeah, it is. Well, so basically all the regulations in the US, you know, environmental, this and that, builds its own sort of stakeholders and lobbying groups. And then it’s in the interest of these lobbying groups to prolong the process so that they can absorb a higher share of the rent. Whereas in China, because they have this command structure headed by the Chinese Communist Party, the secretary of the city or province can cut through all the red tape because he’s the boss of all the regulators,

Dwarkesh Patel: 1:07:25 Mhm.

Victor Shih: 1:07:26 most of the regulators, let’s say at the local level. So if he wants to do something he can cut through all of it

Dwarkesh Patel: 1:07:31 Yeah, yeah.

Victor Shih: 1:07:32 as long as he can benefit somehow himself. So I think that would be, I would say that’s a difference.

Dwarkesh Patel: 1:07:43 Interesting. Okay. And then what would a solution to this problem look like?

Victor Shih: 1:07:50 Ah, okay. So I would say, and of course the Chinese government will never ever in a million years do it in its current form, which is to reduce a lot of the other unnecessary spending, like national defense, like a lot of the industrial policies, and then use the savings to bolster domestic demand with welfare policies and to gradually decrease overall amount of local government debt.

Dwarkesh Patel: 1:08:18 Yeah.

Victor Shih: 1:08:22 That I think is economically sound because it’s like China already has the largest navy in the world, why does it need a navy that’s even bigger than the largest navy in the world? It doesn’t need that. I mean it has more than enough for national defense, it has some of the best jet fighters in the world already, some of the best missile systems. Like no one is going to invade China or come close to it.

Dwarkesh Patel: 1:08:45 Right.

Victor Shih: 1:08:46 So they actually don’t need it. I mean the tech race, I mean I do I’m somewhat sympathetic to China, you know, because now that there’s US policies preventing Chinese companies from buying all the chips and AI so it wants its own capacity. is spending a lot of money on it. So that part I’m somewhat sympathetic but other aspects of industrial policies, you know for some of the battery stuff or solar power, etc, it can certainly reduce a lot of subsidies. It can increase taxes, actually, on many export-oriented firms, which is one of the reasons why there’s a big trade surplus in China and trade imbalance between the US and China. Get more revenue that way to finance domestic demand and to pay down local government debt over time. The Chinese government of course will never do that because it has a very high, places a very high priority over competition with the US across many different fronts and to get dominance over multiple supply chains. But you know in a world of global trade, why do you need that? I mean the same thing goes for the US. The US doesn’t need, you know, dominance over every single supply chain.

Dwarkesh Patel: 1:09:54 Yeah. Why is it the case that economists, so when they’re talking about rebalancing more towards consumption, they’ll often suggest that you should do increase the amount of welfare. But it seems like the more straightforward thing to do would just be to get rid of financial repression so that by default savers would have more purchasing power, get rid of currency devaluation, again the same effect. So why not get rid of the regressive taxes rather than the welfare?

Victor Shih: 1:10:28 Uh, because yeah if you just get rid of financial repression it will only benefit the savers, which is, you know, 10% of households. Net savers, 10-20% of households. If you look at median households they have very little savings actually, besides the home in which they live. So yeah, so there’s this misconception like, ‘Oh, China has like the largest savings deposit in the world,’ and that’s true, but if you look at the distribution of these savings deposits, it’s highly concentrated in the hands of 10%. So even if you say, ‘Oh okay, interest, deposit interest rates suddenly goes up to 4%,’ it’s only going to benefit, you know, the people with large amount of savings. Most people they’re not going to benefit from it. Most people if they don’t have better welfare services, they’re going to save like crazy in anticipation of getting sick. So you do need better, you know, certainly medical care insurance in China. I mean that’s something, you know, frankly I worry about the US. You know if you cut back on Medicaid, it’s going to encourage precautionary savings, which you can say, ‘Well but that’s a good thing, Americans save too little.’ Maybe, but then your short-term consumption’s going to come down.

Dwarkesh Patel: 1:11:35 Okay, so you mentioned a second ago, ‘Well why do they need this big military?’ Obviously the reason is not just self-defense but a potential invasion of Taiwan. So given your reading of elite politics, what’s your sense on the probability that they would actually invade Taiwan this decade?

Victor Shih: 1:11:53 Uh, so first of all, it’s, I think we do know something about Xi Jinping’s preference mapping. When it comes to this issue, he has said a number of times that, you know, obviously China should be united, so I do believe that’s sincere. But clearly his desire to achieve that is not so strong that he would take a very risky gamble achieving it, because if that were the case, he would have done it already. He’s been in power for 12 years. You know, he hasn’t done very much. I mean, certainly the amount of pressure on Taiwan has ratcheted up, and also from his other policy-making, we know that he’s not a reckless policy-maker.

Dwarkesh Patel: 1:12:36 Well, zero COVID was kind of reckless.

Victor Shih: 1:12:38 Well, do you think that was reckless? Well, so that was actually conservative, right? So he preserved the lockdown longer than was necessary.

Dwarkesh Patel: 1:12:46 Yeah. I mean, it depends on how you—but you could say the same thing about Taiwan. It’s like, it’s conservative because in the long run, maybe they’ll pose a national security threat, right? But he is a change from the norm.

Victor Shih: 1:12:58 No, no, but in other words, if he wants it so badly that he would just do whatever it takes to get it, he would have done it already.

Dwarkesh Patel: 1:13:21 Yeah, I guess you could argue that he has been trying to build up enough self-sufficiency that in the aftermath, China would not be left in the lurch with food or power, and that’s why they’re doing clean tech, that’s why they’re doing investments in semis and so forth, so that over the next few years, if they build up enough self-sufficiency, they’re in a position where they actually could, whereas he couldn’t have done the same thing in 2015.

Victor Shih: 1:13:41 Right. So, yeah, so then, then we’ve seen the Chinese government engage in all this behavior, stockpiling, you know, this and that, like oil, grain, so on and so forth, enlarging its navy, you know, building all this landing capabilities and, and so on and so forth. So then, then one argument is like, well, so there’s a threshold. Once he feels that he reaches that threshold, then he’ll go for it. You know, but then for me, it’s like, I don’t think that threshold is fixed, right? It’s really conditional on a lot of different things, and there are some exogenous factors that will increase the threshold, other exogenous factors will lower the threshold. So I would say more recently, one factor that might have increased the threshold for him acting is what happened in Ukraine. Right? So Ukraine, you have a case where Putin was very confident about taking over all of Ukraine within a short period of time. Turns out he was wrong. He was getting all this bad intelligence, so on and so forth. I think for Xi Jinping, he cannot discount that possibility, and that very likely increased the threshold.

Dwarkesh Patel: 1:14:45 Interesting. This gets to the question of how much information flow is there, how much true information is getting up to Xi Jinping? In authoritarian systems, obviously, there’s a danger that the leader is not getting told what they don’t want to know. But you were mentioning a second ago… that you think that the top leadership is in communication with experts all the time, is in meetings all the time where they are learning about what’s happening. Is that your sense of— but then you read about Zero COVID and it seems like by year three it was very obviously a mistake. So to the extent that it wasn’t stopped at that point, was that because information wasn’t getting up to the top leadership? Similar story with, I think there was a very famous case in 2024 where Xi said, why don’t we have billion-dollar unicorn tech companies in China? And people were responding, does he not realize that the 2021 tech crackdown is the reason that’s the case? Did somebody not tell him? So what’s your sense on how much he is getting told things which he might make him uncomfortable? For example, that we might not succeed in Taiwan.

Victor Shih: 1:15:50 Yeah, so this, yeah, goes to a question of the role of expertise in the Chinese government. I think that’s very interesting question. I mean, frankly, I mean, I’m trying to use some kind of data to try to assess that. But my intuition is this. The top leadership always listens to experts. So unlike, you know, some cases in the US, the experts are respected, they’re listened to. But there are sort of interest groups that are close to the leadership, they also know that. So they will present different kinds of experts to tell him different sorts of expert opinion. So sometimes, you know, you would get these conflicting advice from different experts, and he will have to make some kind of gut call.

Dwarkesh Patel: 1:16:34 Mm.

Victor Shih: 1:16:35 Other times, for political reasons, they will just ignore the experts. So they will never say, you know, fake news, we don’t believe in science, and so and so forth. Sometimes even though they know for sure the experts are right, because of other political reasons, they’ll say, okay, you’re right, but we’re not going to do what you want us to do for now, for a bunch of political reasons.

Dwarkesh Patel: 1:16:54 And how did that manifest in Zero COVID?

Victor Shih: 1:16:57 Yeah, so so I think the timing of it. Also the decision not to buy Paxlovid on a very large scale, that might have been some kind of political decision. We show, we have a paper actually, I should send it to you, showing that basically they kind of knew—the experts told them the US vaccines were better, mRNA vaccines were better. But the Chinese government said, okay, we believe you, but still we want to help our domestic pharmaceutical industry, so we’re going to only sell the domestic stuff in China. So that’s very clear. If you look at the propaganda, you know, they began to denigrate the Western vaccines and highlight the Chinese vaccines and so and so forth. But but I think they do meet with experts a lot. Sometimes the experts are manipulated, you know, into saying certain things. Other times the experts are sincere, they listen to them, but they still choose to not, you know, implement what the experts recommend.

Dwarkesh Patel: 1:17:53 Yeah, I guess I’m just trying to get a sense of where does this net out in the competence of the political system. It seems like they have a… track record of making some smart calls but then occasionally making a call that is so bad that it wipes out all the smart calls and zero-covid being an example of that or obviously one-child policy before that not to mention all the things that Mao did. Um and so I’m getting the sense that they will consult with economists or scientists or whatever on some things but then I guess I still don’t understand well then why did they okay let’s we’re not going to buy the American vaccine because of this this is the time of all times to be making sure the pharmaceutical industry in China is in a good position sure but then did they not know that people were locked in their apartments starving and that this was not a long-term viable strategy and so forth. Like how is this compatible with this very technocratic picture we have of the party?

Victor Shih: 1:18:54 Well so so uh they’re technocrats who people who know a lot but um I guess yeah generally speaking protecting the party protecting the state apparatus and uh all the important tools of the Chinese government uh will almost always be a higher priority to uh any kind of technical considerations in a given decision-making. So it’s like well why protect all these pharmaceutical companies? Because they’re state-owned enterprises. They need to survive the ordeal they need to have a national brand name uh and also of course pharmaceutical and new what they call new biology is in one of these industrial plans in order to finance some of these future research in uh advanced biology these companies need cash flows.

Dwarkesh Patel: 1:19:46 What is the succession plan after Xi is gone?

Victor Shih: 1:19:49 Uh there are no plans right now. There are no plans. I mean so the biggest

Dwarkesh Patel: 1:19:52 what’s your sense of what would happen tomorrow he drops dead what happens next?

Victor Shih: 1:19:55 Oh god okay. Uh yeah I think that would be an issue. Yeah if he were to drop dead tomorrow. I mean so one of the biggest impact is this right? So right now financial repression works because so let’s say you’re a billionaire you try to get your money out of China you know you make up some fake paperwork and oh I’m I’m importing a million Rolex watches from Singapore I need to pay an invoice for a billion dollars. Um that order will go from your bank to the state administration of foreign exchange in Shenzhen or in Shanghai and some bureaucrat will have to okay it right large denomination outflows. Right now that bureaucrat is like if I approve this someone in Beijing is going to notice it and I’m going to be in jail within a week. Yeah so I will not approve it. But if that person was like there’s nobody in Beijing. No one’s minding the store and this guy promises me a hundred million dollars if I approve it I’ll go ahead and approve it. So if there’s any sense that there’s no command structure in Beijing even for like a week or two, uh, I think we would have a financial crisis, at least.

Dwarkesh Patel: 1:21:04 That’s interesting. How might you just say because there’d be a tremendous capital outflow?

Victor Shih: 1:21:09 So let’s put it this way, the foreign exchange reserve, everybody’s like, oh my god, so much money, 3 trillion dollars. Uh, right now it’s like less than 5% of money supply in China. So it doesn’t take a panic. Like so even if people want to just rationally reallocate 10% of their assets outside of China, foreign exchange is gone, massive devaluation, etc., etc. You know, huge and they would have to jack up interest rates to like 20% to stop it, then there’s mass bankruptcy, blah, blah, etc.

Dwarkesh Patel: 1:21:41 Yeah. But they can stop it. Is it—

Victor Shih: 1:21:43 If there’s a command structure. Yeah. If there’s a—if people believe that they’re going to go to jail the week after even if they get $100 million in cryptocurrency in Hong Kong or whatever.

Dwarkesh Patel: 1:21:52 And suppose it’s uh, it doesn’t happen tomorrow, but in 2028, 2030, he just goes senile and over the course of months, it’s clear that there needs to be a successor. What does then the succession process look like?

Victor Shih: 1:22:06 Yeah, uh, so there’s historically a way that we know doesn’t work and then there’s a way that could work. Uh, one way that we know historically doesn’t work is to designate someone and it’s like, you’re going to be the successor. I mean, that person’s going to die for sure, you know, in a horrible way. And we’ve seen this for Stalin, we’ve seen this for Mao. Uh, the other way is to have someone that everyone trusts to represent the leader’s opinion, but cannot possibly become the number one person. And Mao had Jiang Qing, his wife. Xi Jinping uh can also do that for his wife or increasingly we see his daughter taking a higher profile. Uh and the reason why these individuals cannot be number one in the party is because of this ingrained sexism in the party that you know, against women. And also they have never held official positions in the party and so on and so forth. Uh, so I think having a transition figure who can basically uh stabilize the whole situation while the different factions fight it out peacefully, hopefully, uh will be a very important ingredient. But sometimes even like these transitional figures are too weak. So Jiang Qing was purged within uh weeks of Chairman Mao dying. Right. Uh, the Long March veterans, they were so close to each other that they worked it out among themselves nonetheless. So that was a pretty peaceful transition besides the purge of Jiang Qing and her colleagues in the Gang of Four. Uh these days, the Long March generation of officials who’ve spent decades with each other, governing with each other, that doesn’t exist anymore. There’s very high turnover, the people who are in the Politburo, they’re different elements in it that they, you know, some of them, a few of them have worked closely with each other over the decades, but many of them worked in such disparate bureaucracies. cies that there is not a lot of trust between them and that probably is by design, right? So you know, you don’t want people on the Politburo to trust each other, to be friends with each other too much because they could get together and usurp your power. Um, so given that there isn’t this social capital that we saw back in 1976, I think this transition is going to be more ruthless, more brutal, and potentially more disruptive.

Dwarkesh Patel: 1:24:27 What are the important factions in Chinese politics today? What is the equivalent of liberal or conservative in American politics? Historically, there’s things like the Shanghai clique or whatever, but what—do those still exist? And if they do, what do they stand for?

Victor Shih: 1:24:44 I mean, they’re all beholden to Xi Jinping in one way or another.

Dwarkesh Patel: 1:24:48 But do they have different ideas about what should happen with China or what should happen with the world?

Victor Shih: 1:24:54 Yeah, generally speaking, you have some people who are more pro-market. These people tend to be people with more governing experience along the coastal provinces like Zhejiang province, Shanghai. And then you have the statist, you know, people, most of their careers in state-owned enterprises. So all the military industrialist, I would guess that they’re more pro-state power. So during, well, but actually those figures, I think will ultimately be somewhat stabilizing. Because you basically saw this in the Soviet Union. For those people, they just want to preserve their vested interest. So it’s like okay, whoever’s in charge, as long as you subsidize the state sector some more, I’ll support you. So they in the Soviet Union after Stalin died or even, you know, between the sort of Khrushchev and post-Khrushchev transition, they were stabilizing politically even though they ended up damaging the Soviet economy by demanding so many subsidies for their own sector. So the fact that you have quite a number of these guys from the military industrial sector as Politburo member potentially could be stabilizing.

Dwarkesh Patel: 1:25:40 What’s your forecast on Chinese growth or the Chinese economy relative to the US if AI is not a thing? Let’s just forecast out everything else in the year 2040. Is it 2x as big as US economy PPP? Is it 3, 4x? Is it 1x?

Victor Shih: 1:25:58 Yeah, no, I think it’s 1x or sort of 1.2x of US economy.

Dwarkesh Patel: 1:26:22 Even PPP?

Victor Shih: 1:26:24 Yeah, well, I don’t quite believe in these PPP calculations.

Dwarkesh Patel: 1:26:30 Oh really? Interesting. How come?

Victor Shih: 1:26:31 Just because the quality of services is very different, you know, in China’s in some cases still worse. In other cases increasingly it’s better than US quality. So I’m not an expert but yeah like even controlling for inflation so real income, overall size of the economy would be similar to the US maybe slightly larger by 2040.

Dwarkesh Patel: 1:26:54 But it’s already 1, 1.2 right? So…

Victor Shih: 1:27:00 No, no, no. Sorry. No, no. Well yeah if it’s not PPP then… disposable income, I think it’s like 70% of US economy, something like that.

Dwarkesh Patel: 1:27:02 So this is a quite bearish forecast because usually people expect it to be quite dominant by 2040.

Victor Shih: 1:27:08 Yeah, because growth rates were slowing down anyway.

Dwarkesh Patel: 1:27:10 Yeah.

Victor Shih: 1:27:11 By and then now you have this trade pressure and so on and so forth. It’s going to slow down. There’s no big consumption stimulus in sight because of the high debt level. So they can’t possibly push growth rates up through that mechanism. So they have to double down on more investment, more supply side. But then at some point is the world really going to switch to 100% buying BYD as opposed to, you know, Volkswagen and you know all these other cars? I don’t know. I think there’ll be a pretty big push back. Or at least some kind of limits, right? So you know, let’s say China takes over most of Latin America, most of Africa, most of Middle East. That’s still not such big markets, right? You ultimately have to conquer Europe, you have to take North America. That’s going to be really tough.

Dwarkesh Patel: 1:27:58 Final question.

Victor Shih: 1:28:00 Sure.

Dwarkesh Patel: 1:28:00 What advice do you have for me on how to land if not somebody on the Politburo, like one step away from the Politburo?

Victor Shih: 1:28:06 Uh, yeah, I think that’s very tough. Anyone who’s been a mid-level or higher official can no longer leave the People’s Republic of China. There are some former vice ministers who can go to Hong Kong. So I think that’s your best bet is like reach out to one of these vice ministers type and to agree to meet with them in Hong Kong or Shenzhen, even better yet. That could be okay if they’ve been out of power.

Dwarkesh Patel: 1:28:31 And you think it could be an interesting interview? Or would they not say anything?

Victor Shih: 1:28:34 Uh, some yeah, it depends on the person. I mean, some people are willing to be a bit more open. But frankly, yeah, there is this atmosphere in China where people are afraid of deviating from the party line because they can really get in trouble.

Dwarkesh Patel: 1:28:53 Yeah, yeah. Okay, this was super useful to get a very much more tangible sense of what is actually how the Chinese political system actually works. We talk about China so much and I just, most of us have such little understanding of even the basic like Congress, president-level Chinese political system. And obviously a lot of it is obfuscated on purpose. So I appreciate you adding a lot more color to it.

Victor Shih: 1:29:18 Great. Thank you for having me.

Dwarkesh Patel: 1:29:20 I hope you enjoyed this episode. If you did, the most helpful thing you can do is just share it with other people who you think might enjoy it. Send it to your friends, your group chats, Twitter, wherever else. Just let the word go forth. Other than that, super helpful if you can subscribe on YouTube and leave a five star review on Apple Podcasts and Spotify. Check out the sponsors in the description below. If you want to sponsor a future episode, go to dwarkesh.com/advertise. Thank you for tuning in. I’ll see you on the next episode.