The AI race won't be won by who builds the best model, but by who can afford to keep the lights on. Demis Hassabis figured that out before almost anyone.
When Hassabis sold DeepMind to Google in 2014, he knew what he was buying: a parent company that prints cash. Every other major AI lab is still scrambling for the same thing. That's the central argument Sebastian Mallaby makes in "The Infinity Machine," a new biography of Hassabis that Axios obtained in advance of its May publication. And unlike most AI mythology, this one has numbers behind it.
Since 2010, Google has invested $12 billion into DeepMind, according to Mallaby's book, as reported by Axios. DeepMind has published roughly 7,000 papers, with three Nobel laureates citing its work. AlphaFold, which solved protein folding, is in active use by over 2 million researchers. These are not benchmark claims. They are operational metrics of a lab that has shipped.
The financial angle is what makes Hassabis different from every other major AI leader. OpenAI is testing ads as it moves toward an IPO, projecting $14 billion in losses for 2026. Anthropic is raising at a $60 billion valuation with no clear path to self-sustaining revenue. DeepMind's parent Google, by contrast, is funding its AI buildout using the lowest percentage of debt among the hyperscalers, thanks to its cash flow from search and cloud. Hassabis doesn't have to answer to investors who want to see a monetization plan. That is not a small advantage when the race is an endurance contest.
Mallaby's most revealing detail: Hassabis and co-founder Mustafa Suleyman once recruited Reid Hoffman to pledge $1 billion to spin DeepMind back out of Google. Lawyers and bankers worked for three years to make it happen. Google held on. The arrangement Hassabis couldn't escape may now be his biggest asset.
The book also surfaces what Hassabis told Axios' Ina Fried at Davos: "We don't feel any immediate pressure to make knee-jerk decisions" on monetization through ads or other formats. Competitors are not in a position to say the same.
The competitive spirit has a downside Mallaby doesn't elide. "The classic criticism of him is that he's so consumed with winning the race and being the person who brings AI to the world is messianic, and it's over the top, and it distorts DeepMind's mission," Mallaby said. Textbook founder behavior, he adds — but it raises real questions about whether the mission or the victory is primary.
DeepMind's Isomorphic Labs, the commercial vehicle for drug discovery, has two programs in Phase I clinical trials. One, partnered with Eli Lilly, is worth $2.4 billion upfront — a structure that suggests Lilly believes in the approach enough to pay in advance for the right to find out whether it works. That is a meaningful bet from a company that has been making drugs for over a century.
What Mallaby's biography makes clear is that Hassabis is not a believer in the techbro theory of change. He is not building AI because he thinks it will make money. He is building it because he thinks it is the most important thing anyone will ever build, and he has spent sixteen years acting accordingly. The world has finally started to agree with him. Whether that makes him right or just early is the question the next sixteen years will answer.
