Silicon Valley is now deeply split over whether Washington should ban Chinese AI models from American soil.
At the center of the storm is Moonshot AI’s Kimi K3, a 2.8 trillion-parameter model released on July 16 that has been matching or beating outputs from Anthropic and OpenAI in long-horizon reasoning and coding tasks.
The accusations and the pushback
White House officials, including Michael Kratsios, have leveled serious allegations against Moonshot AI. The claims: the Chinese startup illegally distilled models from US firms like Anthropic and accessed restricted Nvidia GB300 chips that should never have reached Chinese hands. The threatened consequences include sanctions and potential placement on the Entity List, which would effectively cut Moonshot off from the American tech ecosystem.
Around July 22, more than 200 startups, including members of Y Combinator’s network, signed a letter opposing an outright ban on Chinese AI models. Their argument is straightforward: these models cost roughly 1/20th the price of their American equivalents, often running at $2 to $3 per million tokens.
