On a recent Diet TBPN segment, Jordi Hays and his co-host wandered back to December 2010, when a young neuroscientist named Demis Hassabis raised a small amount of funding for a research outfit called DeepMind. The number Jordi surfaced is the kind that makes you spit out your coffee. He said DeepMind “sold half the company at $5 million post” when it was already, in his framing, one of the most elite AI research labs on the planet.
Sixteen years later, Alphabet (NASDAQ:GOOGL | GOOGL Price Prediction) is a $4.4 trillion company whose AI story runs directly through the lab it eventually bought. So the retrospective question the hosts kept circling is uncomfortable and fair. Did the world’s best AI lab leave a generational fortune on the table by getting into bed with Google when it did?
The $5 million number
A $5 million post-money valuation in 2010 sounds like a seed round for a productivity app, not a stake in what would become AlphaGo, AlphaFold, and the intellectual spine of Google’s Gemini program. Google acquired DeepMind in 2014, reportedly for around $500 million, and the lab has since produced work that, by any measure, changed biology and gaming and shipped straight into the products powering Alphabet’s current results.
Those results are not subtle. Alphabet’s Google Cloud backlog sits at roughly $462 billion, with revenue in the segment compounding at 63%. Q1 FY2026 revenue came in at 21.8% year-over-year growth, and the stock is up 106.01% over the past year. The SEC filing behind the quarter is here. DeepMind’s fingerprints are everywhere in that number.
Did DeepMind sell too early?
Jordi’s co-host said the quiet part out loud. “It seemed like they sold too early,” he offered, and you can see why. OpenAI, a peer lab that stayed independent longer, has been valued in the hundreds of billions. Anthropic just anchored a $1.8 billion, seven-year cloud deal with Akamai. The going rate for a top-tier AI lab in 2026 runs into the tens of billions.
Jordi pushed back. “Hard to say,” he said, before landing a “paper hands Pete” joke at Hassabis’s expense. The co-host then noted a piece of Silicon Valley folklore worth remembering. Back then, VCs were warning founders not to sell to Google, calling it “the bad one”. That framing has aged strangely. As the co-host allowed, “Google’s been very responsible and great, great company”.
Why Google might have been the right call
Consider the case for Hassabis. Frontier AI research eats compute the way a foundry eats coke and iron. Alphabet is guiding to $175 billion to $185 billion in capital expenditures for FY2026, the kind of number a 2014 startup could not have dreamed of raising on the open market. DeepMind inside Google got TPUs, data, distribution, and a patient owner. It also got Gemini, which is now processing 16 billion tokens per minute via API.
Would an independent DeepMind have built AlphaFold if it were burning through venture money worrying about the next round? Perhaps.
The co-host conceded the uncertainty. “Maybe there’d be another path or something,” he said. The segment also touched on Google’s Veo video models, which the co-host described as “so close and yet so far” from indistinguishable-from-reality output. That gap gets closed with compute. Compute gets bought with Google’s balance sheet.
The takeaway for anyone scrolling GOOGL on their phone is a reminder that the acquisitions that look like steals in hindsight often looked like lifelines at the time, and that the counterfactual, an independent DeepMind IPO in 2024, lives only in group chats and podcast segments like this one.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks