Ant's $320 Billion IPO Was Stopped at the Eleventh Hour: Angela Zhang on China's Tech Tightrope

Nathan Labenz opens by admitting he does not know much about China and is trying to catch up fast. A lot of American AI discourse, he says, sits on assumptions about Beijing's technology, its intentions, and its attitude toward the West, and he has a sneaking suspicion that many of those assumptions are not well founded. His guest is Angela Huyue Zhang, a professor at the Gould School of Law at the University of Southern California and author of High Wire: How China Regulates Big Tech and Governs Its Economy. She grew up in China, did her undergraduate degree and some legal practice there, worked on China-related projects overseas, taught for seven years at the University of Hong Kong, and has also lived and worked in Brussels and London.

Asked what would most surprise an average American visiting China, Zhang's first answer is the high-speed train. The United States still does not have one. California has been trying for a long time and, as she understands it, has finished about a third of the project, with no one sure when it will be done. A San Francisco to Los Angeles route, she says, would take China one or two years. She is not claiming China is ten feet tall. She is claiming that on execution, speed, and efficiency, the gap is sitting there in plain sight, and she finds American inefficiency on this kind of high-tech infrastructure appalling.

Labenz flags that some of Zhang's views are controversial and that he chose not to push back on each of them in real time. He is not sold that the DeepSeek moment proves export controls will not matter, and he notes that Jeff Ding's work on technology diffusion points the other way from Zhang's claim that diffusion is a Chinese strength. He asks listeners to take her self-description as a neutral analyst seriously for the length of the conversation, and promises contrasting views on the same feed later.

Fieldwork in China is getting harder for foreign researchers. Zhang's method is a pile of older advantages: she was born and raised there, practiced there, kept industry contacts, and spent seven years in Hong Kong, close enough to accumulate firsthand experience. The book also rests on empirical work, including court cases, plus secondary sources. She compares herself, with a laugh, to an AI: gather everything you can and try to answer the question. The part she will not relativize is presence. It is impossible, she says, to truly understand China without being there, and you do not get it from an interview. You have to be absorbed in the environment until the texture of daily life starts to teach you.

Labenz has never been and is thinking about arranging a trip, both for intuition and as a small example of ordinary engagement. Zhang welcomes him and the listeners. Living in the United States, she sees a lot of fear about traveling to China. The vast majority of people who go, she says, have a good time. It is not as scary as the media portrays, and visitors are often surprised by the technology on arrival. Language is less of a wall than Americans expect. China is a leader in voice recognition, and the tools are built into everyday software. When she lived in Hong Kong, her domestic helper spoke only English and her mother-in-law spoke Chinese; they communicated flawlessly on their phones.

The information gap, in her view, runs mainly one way. China understands the United States much better than the United States understands China, which is unsurprising: the Chinese government is opaque, the bureaucracy is efficient and deliberately ambiguous, and internet censorship makes the interior of the bubble hard to see. Read Chinese news and you hear almost everything happening in America, mostly the bad news. Each side is looking through a filter. She does not want to paint a rosy picture. The Chinese economy is at what she calls rock bottom; young people are pessimistic, perhaps the lowest sentiment in decades; and her book is, for the most part, critical of the governance model. She will still grant the other side of the ledger. An authoritarian regime can get things done quickly and puts more weight on the majority. Western coverage treats cameras and facial recognition as a nightmare. Ask an ordinary Chinese grandmother, Zhang says, and many people do not care, or feel safer, because crime is extremely low. The Communist Party knows how to get endorsement from the majority. Approval was high when China achieved zero Covid and resumed production while the rest of the world looked chaotic. Sentiment turned after Omicron, when other countries opened and China stayed under severe lockdown.

The core of High Wire is a dynamic pyramid Zhang has also used on Covid policy and the property crisis. Tech is simply the richest case study. Four actors sit in the stack: top leadership, meaning Xi Jinping and the rest of the Politburo; regulators and bureaucrats who do the daily work of writing rules and enforcing them; firms, the main objects of regulation; and platform participants at the bottom, the merchants, delivery workers, and consumers who live on the apps. The United States and the European Union have the same four roles. The difference is how they are organized, and the difference that matters most is power imbalance.

Labenz looks up the Politburo and gets 24 names. Are we talking about a couple of dozen people, not a broad elite or a Congress-scale body? Zhang's answer is sharper: you can just think about it as one person, President Xi, because he basically controls the top. Regulators are not independent. They answer upstairs, not to the public. Below them is a huge apparatus, 31 provincial governments plus central ministries, with the expertise to implement law. When they do, they watch the policy winds. In China, she says, the signal is incredibly important. In some ways the signal is the policy, and the policy is the signal. Local officials have no independent power base. They are not elected and not appointed by some rival bloc. Everyone answers to one boss upstairs. They have discretion, and they also have their own bureaucratic incentives, but they need to align what they do with the top's initiatives, or at least frame it that way.

Firms sit below the agencies. In other jurisdictions, companies go to war with regulators. In China they almost never directly confront the government, whether they are state-owned, private, or foreign. Elon Musk, Zhang says, behaves like a sheep in China and tries to please the government. In the United States he sends a poop emoji to an agency. She cannot imagine him daring that in China. At the bottom, ordinary users are weak bargainers against platforms, as they are in the West, and censorship mutes their complaints, until the state wants those complaints heard. Then the machinery amplifies them, because public anger helps agencies claim legitimacy. Zhang calls the government extremely media savvy.

She places Europe in the middle of the spectrum. The European Commission has investigated almost every major U.S. tech firm, and those firms almost always challenge the Commission in Luxembourg. Courts are deferential, but businesses occasionally win, they keep fighting, and officials do not hold grudges, because everyone treats this as how a rule-of-law system works. The United States, in her mapping, is the opposite pole from China: business seems to have the upper hand. Five years of hearings and CEO testimony produced no federal law to discipline Big Tech. The one statute Congress did pass was the TikTok ban-or-divest law. There is still no federal data-privacy statute. The data-flow restrictions that exist point at TikTok or at transfers to adversarial countries, including China. Almost all U.S. tech regulation, she says, targets companies with a Chinese identity, while domestic giants face almost nothing. Tech CEOs surrounding the president on inauguration day, with Musk assisting, makes her worry about the other extreme: crony capitalism.

How the censorship machine actually staffs and updates itself is, Zhang says, opaque to almost everyone except a few insiders. What is not a mystery is the scale of the investment. Both the state and the large platforms pour resources into content moderation. For any company in social or content businesses, political alignment with the Party's socialist values is a first-order compliance job. Platforms receive instructions from the Cyberspace Administration of China about what may be posted, with timely updates as circumstances change. Labenz had assumed platforms might salt a little self-protection into the same pipeline. Zhang's sharper point is that the state can turn public complaint up or down on purpose.

Didi, China's Uber, wanted a U.S. listing. CAC asked it to postpone the IPO and get data-security compliance in order. A ride-hailing giant holds sensitive geographic and personal data; sending materials to American regulators looked, from CAC's standpoint, like a national-security problem. Didi pressed ahead anyway. Two days after the listing, CAC opened a cybersecurity review. The stock tumbled more than 10 percent, a chunk of market cap vanished, and international investors reached for a phrase Zhang repeats: China is uninvestable. The legal tool was a stretch. Cybersecurity checks of that kind, as written, are about procurement, whether a firm is buying insecure equipment, not about filing documents with the SEC. Zhang calls the move dubious, and she thinks the regulators knew the footing was soft.

After the action came the legitimacy campaign. Chinese media began calling Didi a traitor for turning sensitive data over to the U.S. government. Liu Qing, Didi's president and a former Goldman banker, was labeled someone who had sold Chinese secrets. Posts circulated claiming Didi's traffic data could show which ministries were busy at which hours. Ordinary people do not know what a cybersecurity review is, or whether there is a clean legal basis, but the nationalist story is easy to follow. The same script ran after Ant Group's IPO was suspended: a flood of posts calling Jack Ma a greedy businessman building a fintech machine to rip off Chinese people, plus reports of dishonest practices. The media machinery, Zhang says, is part of the bureaucracy. Propaganda, the press, and the regulators work in tandem. Information control can mute complaint in ordinary times and amplify a chosen complaint when an action needs covering fire.

Volatility, in Zhang's vocabulary, means the regulatory pendulum travels a long way. From early 2010 until Ant's IPO disaster in late 2020, consumer tech enjoyed a roughly decade-long honeymoon of light touch. That laxness, she argues, planted later problems. The sector became extremely concentrated, a duopoly of Alibaba and Tencent, while the United States at least has seven major tech firms. Had competition issues been handled earlier, the later crackdown might have looked different. Then came about eighteen months of intense pressure, from the halted IPO into early 2022, after which policy reverted toward normal. AI regulation today, she says, is again really lax. Fragility is the third trait: because everything hangs on signals from the top, a turn in the wind can hit firms and markets before they can react.

Labenz wants to know whether Alibaba and Tencent were designated winners or market winners. Zhang's answer is both-and, with the stress on competition. The Great Firewall worked as a trade barrier. Baidu faces far less search competition than Google would have; Bing has about 1 percent. Even in the counterfactual without the wall, she still thinks domestic firms would have won, looking at TikTok abroad, at Chinese apps filling U.S. download charts, at Temu's prices about 30 percent below Amazon's, at Shein, at DeepSeek. Inside China, though, the fight among local companies was brutal. Why it was Alibaba and Tencent rather than some other Chinese rival is, in her telling, because they won those fights fairly, so to speak, in an extremely competitive market.

Jack Ma's speech was a catalyst. The deeper cause, Zhang says, was tension between financial regulators and Ant. Selling clothes, cabbages, toys, or games does not threaten political stability. Crossing into finance at enormous scale does. A single private entrepreneur who had controlled more than 50 percent of Ant, sitting on the world's largest fintech firm, is a picture the Party finds hard to accept. She also has regulatory sympathy. Ant did all sorts of financial business while insisting it was not a financial institution, changing the name from Ant Financial to Ant Group, and calling itself "tech-fin" in the prospectus. The IPO was valued at $320 billion, higher than JPMorgan at the time. A firm spun out of Alibaba less than six years earlier now looked like the largest financial institution in the world, with an opaque controlling shareholder. Regulators saw a bubble and a systemic risk. The conclusion at the top, as she reconstructs it, was that these companies had entered the core of the economy, so their influence had to be reduced and they had to be reminded who the big boss is. She does not think Ma was under house arrest. Xi met him again recently. The circle turned.

The recent thaw is not a green light for consumer finance. Look at who Xi invited to a recent symposium: people in AI, EVs, clean energy, robotics, the boss of Huawei, and the founder of DeepSeek, whom nobody knew a year earlier. Games and consumer apps can continue because youth unemployment is too high and jobs are desperately needed. That will not transform the economy. Xi's project, in Zhang's reading, is to move China from an old, dirty, property-bubble economy toward high-end manufacturing and clean-energy exports. She thinks China has mostly done what it set out to do, including the controversial Made in China 2025 goals; a Bloomberg report she cites says China hit most of those targets and now leads in many of the technologies it named. She will not yet call the path a success. The direction, and the pressure from U.S. chip controls pushing Beijing toward self-sufficiency, is clear.

Zhang borrows Joseph Schumpeter's cycle of innovation, adoption, and diffusion. China, she says, is not good at 0-to-1. True scientific breakthroughs almost never happen there. The transformer, or Geoffrey Hinton figuring out how to use neural networks as the architecture, is 0-to-1. China is strong at adoption and extremely strong at diffusion: lower the cost, raise the efficiency, put the thing in more people's hands. From 1 to 10, it does very well. That is why she does not treat DeepSeek as a technology breakthrough. It is a demonstration of diffusion.

The part that surprised her was Silicon Valley's reaction time. DeepSeek had already been known in China for more than a year. Around mid-2024 it launched a price war in the AI community, slashing prices until others, including leading labs, had to follow. In China it was called the Pinduoduo of AI, Pinduoduo being Temu's parent and a specialist in deep discounts. Western attention arrived eight or nine months later, as if the firm had appeared overnight. Zhang reads that delay as the same information gap: American companies had little interest in China and underestimated it. Two opposite stories now dominate U.S. commentary. One says China is ten feet tall, already ahead, and must be contained. The other says China is three feet tall and its innovative capacity is overrated. She is trying to be the person in the middle who presents what China actually is.

AI regulation, for now, matches the old honeymoon. There are standards and expectations, Labenz says, but not pre-launch reviews of every product. Firms go to market; if something goes wrong, they can expect to be smacked. Zhang agrees the stance is really lax. In that environment, rapid iteration at a firm like DeepSeek is not mysterious. Unless the United States bans Chinese apps, which she thinks is difficult, she expects more Chinese consumer-tech firms to flood into the U.S. market on the Temu-Shein-DeepSeek pattern: take something that already exists and make it cheaper and more widely available.

Did DeepSeek need Xi's personal sign-off to release an open-weight frontier-level model, or did the leadership hear about it on Twitter with everyone else? Zhang thinks the company was quite independent in turning out its products. Every business in China, foreign or domestic, operates under the government's authority and has to play by Chinese rules. The true creative destruction, though, comes from private firms. She is willing to be harsh: anything the Chinese government touches, it is not doing that well. When the state allocates resources, you get corruption, inefficiency, and loss. The semiconductor industry has produced scandals and limited capability, which is why China still leans on Huawei. Huawei, she notes, is not a state firm. It is a private business. The animal spirits still sit with entrepreneurs.

DeepSeek is a maverick even inside that private story. It is not a Big Tech company. It is a quantitative hedge fund that, because of regulation, was restricted from some financial trading, poured resources into training models, and became an overnight success. Zhang calls that idiosyncratic and hard to copy. It does not mean China has only DeepSeek. Alibaba, Tencent, ByteDance, and a crowd of AI startups have also made real progress on efficiency. DeepSeek is the best; the others are not far behind. She does not think the government created those results. She does think it is supportive of AI development. Asked for China's first priority, she still says hardware, semiconductors, because that is the bottleneck.

Labenz asks about a recently announced mega-fund, more than a hundred billion dollars, sometimes described as a quarter trillion in PPP terms, and about American talk of a Manhattan Project for superintelligence, plus the rumor that monitors now sit in DeepSeek's offices. Zhang thinks DeepSeek will need to communicate closely with the government now that it is a major player. She does not describe an occupying force. Open-sourcing, in her reading, is a long game: a public good that makes the firm more credible, helps China look like the second AI superpower, and lets widely used open models shape standards. The state knows it is not best positioned to create or run the technology. It still wants to guide the technology toward national interest.

U.S. export controls, she argues, may fail in a specific way. They aligned private-sector interest with the state's interest in self-sufficiency. Once firms cannot buy the advanced foreign technology, they have to innovate and work with the government. Washington still has voices, Dario Amodei among the most vocal, calling for smarter, tighter chip controls. Zhang thinks that school of thought ignores unintended consequences, not only DeepSeek's efficiency tricks but that public-private alignment. She is more optimistic than the "decade behind" line on Chinese semiconductors. The country has 1.4 billion people and a lot of capable engineers, including people who returned after U.S. anti-espionage investigations under the China Initiative. At the symposium with Xi, Huawei's Ren Zhengfei said the most important thing is that we have people: even if we lose everything else, as long as we have people, we can innovate. Her projection is another DeepSeek moment in a few years, a holy-cow semiconductor breakthrough, and that is when she thinks the tech war ends, because the remaining obstacle is gone.

When Zhang says the tech war will end, Labenz wants to know whether that means friendship, or whether the two countries have a fundamental clash of interests. She asks to come back for another episode. She is writing something called America's Legal Gambit Against China. The short preview: U.S. public opinion toward China dropped sharply after 2018 and has kept falling. Demand-side causes include the trade shock to manufacturing jobs and China's geopolitical rise. The supply side is her focus. Foreign policy concentrates power in the executive in a way that, she says, makes the United States look a bit like China. Trump made toughness on China a brand, broke norms by invoking Section 301 tariffs, restricted capital and firms such as Huawei and ZTE, moved to ban TikTok, and launched the China Initiative against ethnic Chinese scientists, many of whom left. Tough talk also inflated the demand it claimed to serve. Biden inherited almost all of it and doubled down on export controls. In trying to contain China, she argues, the United States is becoming more like China: concentrated power, volatile practice (Trump as TikTok's "biggest savior"), and fragile outcomes. DeepSeek wiped out more than a trillion dollars of U.S. tech market cap. She asks what the controls achieved.

Does Beijing now see the United States as a long-term enemy, or as a volatile administration that can be waited out? Zhang's answer turns inward. China's biggest enemy is never the United States; it is that they mess themselves up. High Wire could also have been subtitled how China has messed itself up. The tech crackdown wiped out trillions in market cap, scared private capital into calling China uninvestable, and sabotaged entrepreneurial spirit. Tariffs and export controls are obstacles. They are not, in her ranking, the biggest ones. The biggest is whether China can offer vision and a rule-of-law system that gives people confidence to invest. The same sentence, she says, applies in reverse. America's biggest enemy is not China. It is American democracy in a constitutional crisis, America messing itself up. Even the DeepSeek story, read that way, is not only geopolitics. Silicon Valley already knew Chinese firms were good at model efficiency. U.S. giants with abundant resources did not follow. Zhang's Project Syndicate piece argues that U.S. AI is too concentrated: Nvidia as a virtual monopoly with operating margins over 60 percent, four big clouds, vertical integration, profits upstream, and downstream labs that do not make money. The business interest is to keep compute scarce and demand looking crazy. DeepSeek succeeded; the United States, in that telling, fell because concentration blinded the innovation path, not because the talent was missing.

Her own product, she says, is a niche: a scholar who grew up in China, knows Europe, lives in the United States, and tries to be equally critical of both governments. Cass Sunstein once said she had the most elaborate personal website of any American law professor; search Angela Huyue Zhang. On the AI-safety rumor that Xi might be a "doomer," she has no idea what he thinks. What she observes is both countries pressing AI at full steam. If you are doing that and you are rational, you should not be a doomer. She finds the trajectory sad. She has written on the promise and perils of China's AI regulation: a very lax stance can sow later disasters, and the pyramid's recurring theme is that problems do not get top-level attention until it is too late, as with Covid, tech, and property. The United States, she worries, is following that lead, relaxing while remaining the frontier leader.

Track-two dialogues are not what the world needs most right now, in her view. What the world needs is a disaster, contained enough not to kill everyone, ugly enough to show leaders what misuse and accident can do to social and political stability. The biggest obstacle to U.S.-China AI cooperation is not geopolitics. It is that we have not seen the disaster and still do not understand the systems. She quotes Hinton: we do not understand how these things work because we do not understand how brains work. In the Cold War, even at the brink, Washington and Moscow signed nuclear agreements because they knew the weapons could kill. We do not know that about AI, so regulators are not serious. In the meantime she likes the idea of an independent body modeled on the IPCC, staffed with scientists, sharing intelligence on capabilities and feeding governments the same picture. If there is a 1 percent or 0.01 percent chance of losing control, she says, it is worth buying insurance. Walking away from regulation until it is too late is not a wise approach. Labenz does not disagree. Zhang's closing invitation is the same as her opening one: come see China, and do not collapse three different layers of the tech story into one slogan.