Why Can Hassabis Lead Google's DeepMind to Overtake OpenAI?

The turnaround: strategy, science and a product moment

Demis Hassabis (德米斯·哈萨比斯) is the unexpected face of a Silicon Valley comeback. Once eclipsed by OpenAI’s rapid GPT rollouts, Google (谷歌) — through its DeepMind group — reportedly regained ground with last November’s Gemini 3 release. It has been reported that Gemini 3 closed gaps on key benchmarks and in some areas even outperformed OpenAI’s latest models, marking DeepMind’s shift from pure research lab to product contender.

From chess prodigy to AGI architect

Hassabis’s backstory reads like a tech parable: a chess prodigy educated at Cambridge who co‑founded DeepMind in 2010 with a mission to “solve intelligence.” Under his leadership DeepMind produced AlphaGo and AlphaFold — breakthroughs that reshaped AI research and earned global acclaim (some commentators suggested Nobel‑level impact, though Hassabis has not been awarded a Nobel). It has been reported that his ambition to simulate whole environments (the internally named “Gaia” project) proved resource‑intensive and slow to commercialize, prompting a pivot toward large language models that could scale faster into products.

Leadership, scale and geopolitics

So how did Hassabis get DeepMind across the finish line? The answer combines scientific rigor with new managerial muscle: DeepMind has grown from hundreds to thousands of staff and has been folded more tightly into Google’s engineering and cloud infrastructure, giving it access to the compute and deployment pathways needed to compete with OpenAI. It has been reported that Sebastian Mallaby’s new book — and its Chinese translation by 周健工 — documents Hassabis’s leadership evolution in detail. The race is also shaped by geopolitics: U.S. export controls, chip supply dynamics and talent flows all influence which firms can scale AI fastest. In short, DeepMind’s comeback is as much about organizational and infra scale as it is about algorithmic insight.