Last month, Stanford University published research that contradicts the dominant narrative on US-China artificial intelligence (AI) competition. According to millions of human voters, the performance gap between the top US and top Chinese AI models has effectively closed (see chart).
Most importantly, Chinese firms are generating frontier AI capability at a fraction of the costs of US firms. According to Stanford, a 23-to-1 spending gap between US and Chinese private firms is producing a 2.7% performance gap. This has direct implications for US-based AI companies whose valuations rest on the assumption of a structural advantage. It also suggests the DeepSeek shock of January 2025 – when the launch of a single Chinese model briefly erased a trillion dollars from US tech stocks – was not an anomaly, but the first sign of a convergence that had been underway for several years and has continued since.
A key question in the coming years will be how Western investors can profit from China's AI capabilities. On April 27, the Chinese government cancelled Meta's $2 billion acquisition of Manus, a Singapore-based AI startup founded by Chinese engineers. Beijing is signaling that frontier AI capability is now treated as strategic territory in the same category as rare earths and semiconductors – and is therefore closing the offshore "Singapore-washing" route that Chinese tech firms have used for years to reach Western investors.
