China's market rout and the AI arms race bill arrives

Two stories running in parallel this week illuminate the compounding costs of the US-China technological rivalry. First, Chinese markets lost trillions of dollars in value as investors priced in the mounting burden of Beijing’s attempt to compete with the United States in artificial intelligence — an effort requiring vast capital expenditure on chips, data centres, and talent in a sanctions environment that raises costs at every turn. Second, the United States banned vehicles made by China’s BYD from its Kadena military base in Japan, citing what officials described as an “unacceptable” national security risk from embedded connectivity technology. Taken together, the two stories represent the dual-front nature of the technology confrontation: China paying an economic price for the race, while the United States tightens the perimeter of what it considers acceptable Chinese industrial presence in allied territory.

The received wisdom

The standard progressive-globalist analysis of the US-China tech rivalry views both moves with ambivalence bordering on alarm. The market rout in China is, in this framing, bad for everyone — an interconnected global economy cannot sustain a deliberate technological decoupling without costs that eventually flow back to Western consumers, pension funds, and supply chains. The BYD ban at Kadena is, in the same register, an overreaction: BYD makes electric vehicles, not weapons systems, and treating every Chinese industrial product as a potential intelligence asset risks trade-war escalation that will hurt American consumers and allies alike. There is also a sustainability angle: the energy transition requires cheap electric vehicles, and the cheapest EVs in the world are Chinese. Banning them from even military adjacent environments, the argument goes, subordinates climate goals to security theatre.

A different read

The security concern about connected Chinese vehicles is less paranoid than the globalist critique suggests. Modern EVs are sophisticated data-collection platforms — they carry cameras, microphones, GPS systems, and cellular connectivity, and they communicate with manufacturer servers. The question of whether BYD vehicles parked at a US Air Force base in Okinawa might transmit operationally sensitive data — aircraft movements, personnel patterns, network signals — is not a theoretical one. The US Department of Defense has been consistent in its view, applied first to Huawei telecommunications equipment and now to connected Chinese vehicles, that the potential for data collection outweighs the convenience and cost benefits. Whether that calculus is correct in every case is debatable; whether it is unreasonable is a harder argument to sustain.

The Chinese market rout is the more structurally interesting story. The Guardian’s reporting points to investor concern that Beijing’s AI ambitions are consuming capital at a rate that cannot be sustained without either major external investment — which sanctions make difficult — or domestic financial repression that diverts household savings into state-directed technology projects. This is precisely the dynamic that some Western economists warned about when predicting that China’s model of state-directed technological competition would generate systemic inefficiency. The Soviet Union’s attempt to compete with American technology in the 1970s and 1980s produced genuine advances in specific military domains while imposing an economic burden that contributed to the system’s eventual collapse. The analogy should not be pushed too far — China’s economy is far more market-integrated than the late Soviet economy — but the basic tension between state-directed capital allocation and efficient innovation is structural, not accidental.

There is a third dimension that deserves attention: the question of what Western AI firms are actually doing with their capital. Anthropic’s blockbuster IPO this week — described by BBC Business as a record share listing accompanied by warnings about “catastrophic risk to humankind” — illustrates the paradox at the heart of the American approach. The US is winning the commercial AI race partly by allowing its most capable AI labs to raise capital from sovereign wealth funds, including Gulf states with their own geopolitical agendas, on the basis of technology that the same firms acknowledge may pose existential risks. The implicit bargain is that the United States would rather be the country that produces a potentially dangerous superintelligence than cede that ground to China. That may well be the right call. But it requires a level of civilisational self-confidence in American institutions — their ability to govern this technology once it exists — that the current dysfunction in Washington does not obviously justify.

The BYD ban and the market rout are, in this light, not separate stories. They are two faces of a competition whose full costs and risks are only beginning to be priced in — by investors in Shanghai, by policymakers in Washington, and by everyone else who will live with the consequences of decisions being made under conditions of strategic urgency and incomplete information.

What to watch

Track whether the BYD ban at Kadena spreads to other US military installations in the Indo-Pacific — if it does, it signals a systematic policy rather than a single-base decision, with significant implications for US-Japan trade politics. Watch Beijing’s response to the market losses: if the government intervenes with a major stimulus or state-purchasing program to arrest the decline, it will confirm that the economic costs of the AI race are already politically sensitive. Monitor the Anthropic IPO’s reception in Washington — whether Congress or the executive moves to impose conditions on AI firms raising capital from foreign sovereign wealth funds will indicate how seriously the national-security community is taking AI capital flows. And observe whether any NATO ally besides the United States moves to restrict Chinese EVs from defence environments — coordination here, or the absence of it, will reveal the strength of the Western technology security consensus.


— J