Hassabis's AI belief — and why he once lagged behind OpenAI

Hassabis’s ambition and the Google bet

Demis Hassabis has long styled himself more as a missionary for artificial general intelligence (AGI) than a conventional Silicon Valley CEO. He views AGI as a way to tackle medicine, energy and climate at scale — not merely as a product. Why sell DeepMind? Because the math was simple: AGI demands massive, sustained compute and funding. Hassabis told potential suitors he was tired of fundraising and wanted resources to pursue hard scientific work. In 2014 Google bought DeepMind for about $650 million, giving the London lab access to the “printing press” of ad-funded compute it needed.

A different technical faith — and a missed wave

DeepMind’s technical culture favored reinforcement learning, embodied intelligence and neuroscience-inspired approaches over pure engineering playbooks. That meant the lab led the field on game-playing systems (AlphaGo, AlphaZero) and protein folding breakthroughs (AlphaFold), but it did not immediately pivot to large language models (LLMs) after the Transformer paper in 2017. At the same time, OpenAI’s engineers treated scale and engineering momentum as the path to powerful models. The result: when ChatGPT exploded in 2022, OpenAI captured the public imagination and user growth DeepMind had not prioritized. It has been reported that Hassabis at one point explored raising roughly $5 billion to buy DeepMind back from Google; reportedly Reid Hoffman offered $1 billion and the team even approached Alibaba (阿里巴巴) co‑founder Joe Tsai, who declined.

Aftermath and geopolitical context

The episode reshaped the industry. OpenAI’s product-led surge forced Google to reorganize, integrating Google Brain and DeepMind and elevating Hassabis’s remit; Google’s Gemini later emerged as a leading foundation model. The story also illustrates a wider tension in AI: should breakthroughs be driven by concentrated corporate compute, open research, or long‑horizon science? That debate now plays out against geopolitics, trade restrictions and heightened regulatory scrutiny of dominant cloud and chip providers — factors Western readers should know shape where and how cutting‑edge AI gets built and deployed.