DeepSeek's R1 team came from China's 'genius' pipeline: FT

The stat that pins the story is one worth checking against the feature itself: roughly 100,000 gifted teenagers are pulled out of ordinary schooling in China every year and put through what the Financial Times describes as an ultra-competitive talent stream, a system that traces back to the mid-1980s when China first sent students to the International Mathematical Olympiad. Zijing Wu's debut FT Magazine feature argues this pipeline, not returning US-trained researchers, is where China's frontier AI engineering now comes from.

The evidence Wu marshals reads like a who's-who of Chinese tech. According to the FT, the founder of ByteDance, the leaders of Taobao and PDD, the billionaire behind Meituan, the brothers who started Nvidia rival Cambricon, and the core engineers behind large language models at DeepSeek and Alibaba's Qwen all came through the system. The most concrete claim: DeepSeek's research team of more than 100, the group behind last year's R1 reasoning model that shipped at a fraction of the cost of its US rivals, was almost entirely composed of genius-class alumni.

The macro number is what gives the anecdote its weight. China now graduates around five million STEM majors a year, compared to roughly half a million in the United States, the piece reports. In 2025, 22 of the 23 students China sent to the International Science Olympiads returned with gold medals. Put together, the argument is that debates over any one founder or lab departure miss the structural point. The bench behind the bench is very deep.

The honest caveats are the ones the reporting can't settle. Feeder-program alumni counts are easy to celebrate and hard to attribute cleanly; a founder who happened to attend a competition class is not proof the class made the company. The piece also does not tell you what share of the cohort is quietly leaving for US labs, how the state's grip on the program shapes what alumni are willing to build, or whether the intense early screening throws away kids who would have been better late-blooming researchers than teenage prize-winners. Take the specifics as reported, not settled.

What it does change is the assumption Western AI labs and policymakers have been operating under for a decade. If the interesting Chinese engineers no longer need a Stanford PhD to be world-class, the leverage US visa policy and US university admissions once had over the AI race gets smaller, and the story of who can staff a frontier lab quietly gets more crowded.