Hassabis Says AI Layoffs Are Dumb. History Says He Might Be Right — Or Not.

Demis Hassabis has a different theory about what AI means for tech workers.

The DeepMind chief — who won a Nobel Prize in Chemistry in 2024 for using AI to solve protein folding, then released the result free to the world's scientists — spent Google's I/O conference this week making a recruiting pitch disguised as a philosophical argument. If AI tools genuinely make engineers three or four times more productive, Hassabis's position is simple: you hire more of them, not fewer.

"I have a million ideas, from lab drug discovery to game design," Hassabis told Wired from the conference. "I would love to have some free engineers to go and do those kinds of things."

That logic puts him directly at odds with companies currently cutting headcount citing AI-driven efficiency. Meta, which cut roughly 8,000 employees starting May 20 as it redirected $115–135 billion toward AI investments, has not responded to questions about its hiring strategy. Block, the payments firm led by Jack Dorsey, laid off 40 percent of its workforce — roughly 4,000 people — citing AI-driven efficiency. Amazon and Salesforce have made similar moves, according to Wired. Across the tech sector, companies have cut 100,000 jobs in 2026 while committing $725 billion to AI development.

The two positions are not equally defensible. The efficiency argument — that AI lets the same revenue be produced with fewer people — is internally consistent. If output per engineer triples, a company can cut headcount and maintain revenue. Hassabis is not denying the math. He is arguing the conclusion is wrong: more productive engineers mean more things get built, and more things that get built means more engineers are needed to pursue the ideas that flow from them.

"Perhaps there is an ulterior motive for putting those messages out — raising money or whatever," Hassabis acknowledged to Wired. DeepMind's own constraints complicate the pitch. Budget cycles, government security clearances for some projects, and internal lab politics all shape which hires actually happen. Whether a lab with those constraints can move fast enough to absorb displaced workers before other labs do is an open question.

The historical precedent Hassabis is invoking supports his case — with an important qualification. In 2001, the dot-com crash scattered talent from hundreds of failed startups across Silicon Valley. Google, still private and far smaller than it would become, used the moment to hire experienced engineers at scale — according to the company's own public history and contemporary tech reporting, post-crash hires drove much of Google's growth through that period, contributing to products that defined its next decade. The mechanism worked because Google was small enough and growing fast enough to absorb thousands of engineers at the exact moment supply was abundant. That is structurally harder to repeat when the company doing the hiring is a large, established lab with procurement processes, security clearance requirements, and internal budget politics.

What to watch next is whether DeepMind and its research-peers are actually moving fast enough to absorb displaced engineers in a way that validates the redistribution thesis. As of publication, no public hiring surge at DeepMind was visible in job posting data, and engineers from the named companies have not yet posted visible LinkedIn moves to DeepMind or comparable labs. The dot-com precedent worked because Google was small and fast-moving; DeepMind is neither. Whether that distinction matters — whether a large, constrained lab can still capture the redistribution benefit that a small, fast-growing one once could — is the unresolved question. If it cannot, Hassabis's argument is right about the principle but wrong about his own ability to act on it.