A potential US ban on Chinese open-weight artificial intelligence (AI) models could cost American businesses up to US$12 billion per year, according to calculations by a US-based academic, as technology firms increasingly turn to cost-efficient Chinese solutions. While the exact economic toll of a ban remains difficult to quantify, usage data from New York-based OpenRouter – a large language model (LLM) aggregator – offers a glimpse into the potential fallout, said Daniel Yue, an assistant professor at the Georgia Institute of Technology’s Scheller College of Business. If OpenRouter users were forced to migrate from Chinese open-weight models to top proprietary alternatives, they could face an additional annual bill of about US$2 billion, according to Yue. The estimate was based on token usage and price gaps between open and closed models recorded from July 21 to 27, he said. Extrapolated to the broader US economy, the cost increase could range between US$3 billion and US$12 billion, depending on the country’s overall reliance on open-weight models from China, Yue added. The researcher stressed the figures were an “order of magnitude” approximation rather than a definitive projection, citing the difficulty of tracking usage outside centralised platforms. New York-based OpenRouter, which enables developers to switch between various AI models through a unified application programming interface (API), captures only a fraction of the global LLM inference market. Washington’s anxieties over Chinese open-weight models reached new heights in late July following the release of Kimi K3, a model from Beijing-based Moonshot AI whose capabilities rival top proprietary US offerings from Anthropic and OpenAI on some benchmarks. The launch reignited the Donald Trump administration’s efforts to ban foreign open-source models, Axios reported. US officials later alleged that Moonshot AI had infringed on US intellectual property. But the potential crackdown has triggered pushback from major American tech firms, with industry giants including Nvidia, Palantir and Meta Platforms signing an open letter urging the US government not to restrict open-weight models and warning that “premature restrictions” would “stifle competition or drive innovation overseas”. The economic disruption remains uncertain partly because it is hard to predict whether US firms would switch to closed models or abandon certain AI workflows altogether, according to Yue. Jaya Gupta, a partner at US venture capital firm Foundation Capital, believes the latter is likely. In an essay titled “AI’s 2008 Moment” last week, he wrote that critical American systems and many of the nation’s industries relied on open-weight models to run AI locally. An outright ban could cause a vast portion of AI demand to “disappear quickly”, potentially triggering a collapse in the heavily leveraged AI infrastructure market, where data centres are being built on debt in anticipation of future demand, Gupta said. For many US start-ups, Chinese models have become a vital tool. Ben Cera, founder of AI agent firm Polsia, said in a social media post last week that switching to Chinese open-weight models helped his company slash its monthly AI bill from US$1.2 million to US$100,000. However, not everyone agrees that a ban would cause systemic economic pain. “Today, most US enterprises don’t use Chinese open-weight models at meaningful scale,” said Steve Hou, head of research at AI market analytics platform Silicon Data. “There’s some usage, but not enough that removal would produce a clean, measurable cost [increase].” Hou argued that the real impact would be indirect, pointing to the way open-weight models had driven down the prices of proprietary frontier models. In a social media post on Friday, OpenAI CEO Sam Altman announced an 80 per cent price drop for the company’s lightweight GPT-5.6 Luna model, alongside a 20 per cent discount for the mid-tier GPT-5.6 Terra. Global spending on LLM inference dropped from US$2.07 per million tokens in early June to US$1.67 in early July – a trend possibly driven by the adoption of cost-effective Chinese open-weight models, according to a Goldman Sachs report citing Silicon Data. However, Hou noted that the price advantage of Chinese models might be narrowing. Moonshot’s flagship Kimi K3 is priced similarly to mid-tier US models, meaning a business’ final AI spending also hinges on a wide range of factors such as enterprise discounts and task complexity, he said.