A Tsinghua Talent Pipeline Is Fueling China's AI Boom
A Wall Street Journal analysis of China's AI rise zeroes in on one throughline: a Tsinghua University talent pipeline that has been years in the making.
Zhipu AI founder Tang Jie and Moonshot AI founder Yang Zhilin were mentor and student in a Tsinghua lab more than a decade ago. Tang supervised Yang's machine-learning research and even recommended him for Tsinghua's top academic prize. Today, each runs one of China's most prominent AI model startups — Zhipu AI and Moonshot AI — and both have become faces of the country's AI startup wave.
Their stories didn't start with DeepSeek's breakout. Tang has been researching machine learning for about 25 years. Over the past two decades, Chinese universities produced a generation of computer scientists, and the internet era piled up huge amounts of data and engineering experience. Some researchers later joined American tech companies, then came home to found startups, connecting academic research, engineering, and commercialization.
Tang set his sights on AGI — artificial general intelligence — early. After GPT-3 was released, he said he wanted to build a machine that 'thinks like a person.' In 2019, he spun his Tsinghua lab technology into Zhipu AI, and he's been betting on large models ever since.
China's AI companies are chasing the US giants on a different route. With chip and funding constraints, they can't replicate OpenAI and Anthropic's high-investment playbook. Instead, they're hunting for breakthroughs in algorithmic efficiency, open source, and engineering optimization. DeepSeek's MLA and MoE techniques are a prime example of that 'fewer resources, more efficiency' mindset.
For the past few years, China's AI competition has looked like a model race. Underneath, it's the payoff from years of talent accumulation. Research that once stayed in the lab is now flowing into industry through a wave of startups.