When cheap AI becomes a secret weapon

There's a quieter fight happening beneath the usual headlines about who has the smartest chatbot. It's about price. Sam Bresnick, a research fellow at Georgetown's CSET, laid out the stakes in a Politico newsletter segment, and his core point is blunt: Chinese AI models have gotten cheap enough, and good enough, to start eating into the business models that American AI companies depend on.

For years the assumption in Washington was that the U.S. held a durable edge because its labs — OpenAI, Anthropic, Google — could out-invest and out-engineer everyone else, China included. Export controls on advanced chips were supposed to widen that gap further. But Bresnick's read on the current moment suggests the strategy is running into a problem nobody fully priced in: capability is only half the battle. Cost is the other half, and China is winning that half.

His phrase for it is worth sitting with. He says a Chinese model is now "considerably cheaper and almost as capable" as its Western rivals, which he argues "fundamentally threatens the business model of the proprietary developers." That's not a claim about raw performance benchmarks. It's a claim about economics. If a company can get 90 percent of the quality at a fraction of the price, plenty of buyers — governments, enterprises, developers building apps on top of these models — will take the discount. And once that shift starts, it compounds, because usage data and revenue feed back into who can afford the next round of training runs.

The broader implication Bresnick is pointing at is that the U.S.-China AI competition won't be decided purely in labs chasing frontier capability. It'll also be decided in pricing sheets and API costs, in which models get embedded into products used by millions of people who never think about who built the underlying system. Chip restrictions can slow China's access to compute, but they don't stop a lab from optimizing hard for efficiency when it has to work within constraints. In some ways, the pressure of doing more with less may be sharpening China's cost discipline rather than crippling it.

What happens next probably depends on whether American labs treat this as a wake-up call or a blip. If cheaper, nearly-as-good alternatives keep spreading, the proprietary pricing model that OpenAI and its peers have leaned on gets harder to defend, especially outside markets where brand loyalty or data sovereignty concerns work in their favor.