Washington and Beijing Now Pick AI Winners the Same Way

China's planning agency has a phrase for how it handles artificial intelligence. The National Development and Reform Commission, the body that decides where state capital flows, announced it would curb wasteful AI investment and disorderly competition, meaning the state picks which firms get funded and which get crowded out ^[1]. Beijing steers money toward preferred startups, and the rest of the market reads the signal ^[2]. The official in charge put it plainly.

We will resolutely avoid disorderly competition or a ‘follow-the-crowd’ approach. — Zhang Kailin

Washington has a word for this: state control. Now look at what the Pentagon did in May. It signed eight AI deals with Google, Microsoft, AWS, Nvidia, OpenAI, and others, explicitly to end its reliance on Anthropic, the one major lab that refused to let its model run autonomous weapons ^[3]. The Defense Department was picking winners too, and the criterion was willingness to enable lethal automation. One government calls this industrial policy. The other calls it deregulation. The American civilian track collapsed on both ends. In November, the administration tried to preempt every state AI law with a single federal standard; the Senate killed it 99 to 1 ^[4]. In May, the president canceled the federal order that would have vetted AI models for safety, after Elon Musk, Mark Zuckerberg, and David Sacks lobbied against it ^[5]. The space where civilian rules might have lived was emptied: a failed centralization on one end, a successful lobbying campaign on the other. In the same window, the military track filled it. The eight-vendor procurement. Project Maven, the Pentagon's AI targeting system, generating live targeting data (locations, elevations, weapon pairings) for strikes on Iran ^[6]. A $500 billion defense increase proposed two months into the conflict ^[7]. The military AI trajectory did not begin with the war. Defense Secretary Hegseth announced an "AI-first warfighting force" in January, a month before the first strikes ^[8]. The war didn't create the machine; it gave it a budget and a proving ground. The conflict accelerated what was already running. Geoffrey Hinton, the AI pioneer, explained why this track feeds itself. The thing that stops rich countries invading poor countries, he said, is their citizens coming back in body bags.

The thing that stops rich countries invading poor countries is their citizens coming back in body bags. — Geoffrey Hinton

Autonomous weapons lower the political cost of war, which makes the military AI track self-reinforcing: the more it works, the more reason to fund it, the more it works. And yet American officials frame all of this as a moral race against China ^[9].

We are leading China by a lot. — Donald Trump

But the machine Washington has built is China's machine. A state directing AI development through handpicked firms, setting the terms of deployment, picking winners by political criterion. Beijing runs it through a civilian planning agency. Washington runs it through the Pentagon. The labels have quietly swapped. The country that spent years warning about state control is now the one demonstrating how it's done.