Who left Google DeepMind in 2026?

Azati highlights that in August 2026 Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le left to start Discovery Loop, with Alphabet as investor and cloud partner. Around the same time Demis Hassabis became Chair of Google DeepMind and Chief Scientist of Alphabet, while John Jumper had left for Anthropic.

Can institutions own their AI breakthroughs?

Azati points out that institutions don't own their breakthroughs — the people who make them do, and those people can walk. Compounding talent turns out to be far harder than compounding compute.

What should companies take from DeepMind's talent losses?

Azati reads the departures as a reminder that being the best lab doesn't guarantee keeping the people who made it best. Retention of individual judgment and track record is its own strategic challenge.

DeepMind May Be the Best Lab. Can It Keep the People Who Made It One?

Back to quick insights

September 11, 2026

DeepMind May Be the Best Lab. Can It Keep the People Who Made It One?

Baryslau Yaravy

AI Engineering Practitioner, Azati

Quick answer: In August 2026, Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le left Google to start Discovery Loop (a venture aimed at automating scientific and engineering discovery), with Alphabet as an investor and cloud partner. The same month, Demis Hassabis stepped back from running DeepMind day-to-day to become Chair of Google DeepMind and Chief Scientist of Alphabet. Earlier, Nobel winner and AlphaFold co-creator John Jumper left for Anthropic; Noam Shazeer had joined OpenAI; David Silver founded his own lab, Ineffable Intelligence. The symbolic kicker: all eight authors of "Attention Is All You Need" — the paper behind modern AI — have now left Google.

The shallow lesson: "DeepMind is losing, it's falling behind"

A tidy narrative that misreads the problem. Capability didn't walk out the door — people did.

The real lesson: capability and retention are different problems

Being the best lab does not guarantee you keep the people who made it the best. The question was never whether DeepMind is a world-class lab; it's whether it can hold on to the individuals whose judgment and track record created that edge.

The sobering part

Institutions don't own their breakthroughs — the people who make them do, and those people can leave, taking the next breakthrough with them. Compounding talent turns out to be far harder than compounding compute.

Full disclosure

Azati is an AI engineering shop, and the author is one of our practitioners — so we watch these moves closely. Talent concentration and dispersion reshape the entire field we build in, and we have a stake in where the best people land. See our AI engineering work.

FAQ

Why does it matter that the "Attention Is All You Need" authors all left Google?

They co-authored the transformer paper that underpins modern AI. Their collective departure is a signal about retention – not about whether the research itself was strong.

Is a talent exodus the same as a lab declining?

No. A lab can stay world-class in raw capability while still struggling to keep the specific individuals whose experience and instincts produced its edge in the first place.

Thinking about AI talent and delivery capacity?

Let's talk

Adapted from a LinkedIn post by Baryslau Yaravy, AI Engineering Practitioner at Azati – read the original. Sources: TechCrunch and GeekWire reporting on the Discovery Loop launch and the Google DeepMind leadership change, August 2026.