Jeff Dean spent 27 years at Google. He built its search index, then its earliest neural networks. On Aug. 5, 2026, he walked out, and three of Google's top AI researchers walked out with him1.
Nine senior scientists left Google DeepMind and Google Brain in 2026. Dean's group makes four. Five more joined three separate rivals earlier in the year. Google calls the pattern ordinary7.
Four Rivals in Seven Months
Google DeepMind's year started quietly. David Silver led the lab's reinforcement learning research and wanted that method to replace language models as the road to smarter AI. Google informed staff of his move in January. By Jan. 16, Silver had a new title: director of Ineffable Intelligence, a London startup built around his bet3.
Summer brought the biggest break. Noam Shazeer, Gemini's co-lead, announced his move to OpenAI on June 174. Two days later, on June 19, John Jumper followed him out the door. The DeepMind vice president was bound for Anthropic, after nearly nine years at the lab5.
By June 24, Jonas Adler and Alexander Pritzel, two more Gemini researchers, confirmed they were leaving for Anthropic too6.
August dwarfed it all. Dean left with Sanjay Ghemawat, a Google senior fellow and his partner of more than two decades. Oriol Vinyals, a DeepMind vice president, and Quoc Le, a Google Brain co-founder, went with them2.
The four announced Discovery Loop the same Wednesday morning, built to automate scientific research from hypothesis to result1. Demis Hassabis stepped back from running DeepMind the same day. He moved to chairman and Alphabet's chief scientist instead.
DeepMind Loses 9 Senior Scientists in 7 Months, 4 on One Day
Confirmed departures of Google DeepMind and Google Brain researchers, by announcement date, January to August 2026
Each pin marks one confirmed departure date; the magenta pin is Aug. 5, when four researchers left together to found Discovery Loop, and the gray pins are the five who left alone or in pairs earlier in the year.
Sources: MediaPost [1]; GeekWire [2]; Fortune [3]; CNBC [4]; Bloomberg (via Taipei Times) [5]; TechCrunch [6]; CNBC [9]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Date | Event |
|---|---|
| Jan. 16, 2026 | Silver joins Ineffable. David Silver, DeepMind's reinforcement-learning lead, named director of Ineffable Intelligence |
| June 17, 2026 | Shazeer exits to OpenAI. Noam Shazeer, Gemini co-lead, announces his move to OpenAI |
| June 19, 2026 | Jumper joins Anthropic. John Jumper, AlphaFold creator and Nobel laureate, posts his departure |
| June 24, 2026 | Adler and Pritzel exit. Two more Gemini researchers confirmed bound for Anthropic |
| Aug. 5, 2026 | Four quit with Dean. Dean, Ghemawat, Vinyals and Le leave together to found Discovery Loop |
| Aug. 12, 2026 | Kavukcuoglu takes over. Koray Kavukcuoglu confirmed as SVP overseeing Gemini development |
Koray Kavukcuoglu absorbed the open seat. Google's longtime tech chief became senior vice president on Aug. 12, reporting straight to Sundar Pichai and running Gemini himself9. Nine names had left the org chart in seven months. One name now covers the gap four of them used to fill.
What Each One Took With Them
Jumper's departure cut deepest at the lab's identity. He and Demis Hassabis shared the 2024 Nobel Prize in Chemistry for AlphaFold, the system that predicts how proteins fold5. Jumper called DeepMind "a special place" on his way out the door, after nearly a decade building that system there10.
Anthropic now owns that legacy twice over. Jonas Adler had worked on Google's coding push. Alexander Pritzel had led pretraining, the stage where a model first learns from raw data, before a human ever fine-tunes it6. Three of DeepMind's AlphaFold veterans now sit inside one Anthropic building8.
Dean frames his new venture as freedom. "One of the beauties of software is that small groups of people can build things that have enormous impact in the world," he told new computer science graduates at the University of Washington in June, before Discovery Loop even had a name2.
Discovery Loop is a public benefit corporation, a company structure that lets its founders weigh a mission alongside profit. Google is a founding investor and its cloud provider, backed further by Radical Ventures and Khosla Ventures2. Silver's Ineffable Intelligence chases that same freedom, built around reinforcement learning in place of the large language models Google still sells3.
4 of DeepMind's 9 Departing Scientists Chose Discovery Loop
Nine senior Google DeepMind and Google Brain researchers who left in 2026, one cube per researcher, grouped by destination
Each cube is one researcher who left in 2026; the label under each group names the lab or startup that researcher joined.
Sources: MediaPost [1]; GeekWire [2]; Fortune [3]; CNBC [4]; Bloomberg (via Taipei Times) [5]; TechCrunch [6]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value | Note |
|---|---|---|
| Discovery Loop | 4 | Jeff Dean, Sanjay Ghemawat, Oriol Vinyals and Quoc Le, all in August |
| Anthropic | 3 | John Jumper, Jonas Adler and Alexander Pritzel, all in June |
| OpenAI | 1 | Noam Shazeer, in June |
| Ineffable Intelligence | 1 | David Silver, named director in January |
| Whole | 9 |
The Case for Leaving: Equity and Small Teams
Money explains part of the pull. OpenAI and Anthropic were both racing toward public offerings this year. A researcher who joined before that listing could hold equity a public company rarely grants so early6. Pre-IPO stock was the one thing only a still-private rival could offer.
Gil Luria, an analyst at DA Davidson, named the other half of the pull. "There is so much demand for limited AI research talent that the frontier AI research labs are willing to do whatever it takes to add them," he said. "This puts OpenAI and Anthropic at an advantage over large companies such as Google because they can promise less bureaucracy and a more focused effort on pursuing Superintelligence," Luria added5.
Shazeer's own exit note backs that read. "It was a difficult decision to move on. I'm incredibly proud of the amazing team at Google and everything we've built together," he wrote, praising the team he was leaving rather than naming a grievance against it4.
Fortune's talks with current DeepMind staff added a harder edge. One engineer described colleagues "hopping over to pre-IPO labs" hoping equity would make them rich. A safety researcher resigned over a Pentagon contract, and three more engineers described 60-hour weeks chasing a delayed model8. Equity, mission and hours all point the same direction, toward a smaller room with a bigger stake in it.
Google's Case: The Bench Still Holds
Hassabis rejects the idea that losing names means losing ground. "We have by far the biggest and broadest research bench of any of the labs out there," he said. "We win our fair share of the top talent," Hassabis added7.
A Google spokesperson echoed him in blunter terms. "The small number of departures won't impact the company's overall trajectory," the person said, framing nine exits against a Gemini team Sundar Pichai has called several thousand people strong7.
Pichai made his own case for Dean and Ghemawat even as he thanked them for leaving. Their work, he wrote, "helped to drive some of the most significant technology transitions, from our early search infrastructure to the neural networks that helped create the modern AI era"1.
Sebastian Mallaby, who wrote a biography of Hassabis, called the panic overdone. "It's a big team. It's several thousand people on the Gemini team, if you count everybody," Mallaby said. "And so we shouldn't over-index on a few famous names," he added9.
Micah Hill-Smith offered a tighter read. His firm ranks frontier models on its own tests. By his count, Google led the chart once this year. It then fell behind rivals on coding tasks, the same ground several departing scientists had worked8.
What to do
Track where a departing researcher lands. Where they land carries the real signal. A pre-IPO lab offers a pay structure only a private firm can match. Founding a public benefit corporation instead points to a mission narrower than the old employer's roadmap.
Weigh Google's own numbers before assuming weakness. Pichai sizes the Gemini team at several thousand people, and nine exits are a thin slice of that. A recruiter pitching against DeepMind should still expect compute and reach few rivals can match.
Watch what follows the exit. Koray Kavukcuoglu now owns that test: ship Gemini 3.5 Pro, then close the coding gap. How he does decides whether 2026's departures read as a blip or a turning point.
By the numbers
- Nine senior researchers left Google DeepMind and Google Brain across 2026, for four separate ventures13456.
- Four of them, Dean, Ghemawat, Vinyals and Le, founded Discovery Loop together on Aug. 52.
- Three AlphaFold veterans, Jumper, Adler and Pritzel, now work at Anthropic8.
- Jeff Dean's tenure at Google ran 27 years before he left to lead Discovery Loop2.
- John Jumper left after nearly nine years building AlphaFold at the lab5.
- Alphabet shares fell about 4 percent the day Dean's departure and Hassabis's reshuffle were announced2.
- Google's shares tumbled as much as 7 percent in June after Shazeer and Jumper's exits became public7.
- Gemini 3.5 Pro missed three release deadlines near the departures8.
What to watch
Watch whether Discovery Loop, Ineffable Intelligence or an Anthropic hire publishes an independent result. It needs to beat Google's own models on a task researchers use every day.
Gemini 3.5 Pro's shipped release date matters just as much. Measured against Kavukcuoglu's new timeline, a working model answers the exodus better than any statement does. A fourth or fifth senior departure before year end would make Google's case harder to sustain.
Sources
- Laurie Sullivan, "DeepMind Chief Scientist Leaves To Start A Company Google Will Invest In," MediaPost, Aug. 5, 2026, https://www.mediapost.com/publications/article/417036/deepmind-chief-scientist-leaves-to-start-a-company.html
- Todd Bishop, "The startup idea that convinced a UW computer science legend to leave Google after 27 years," GeekWire, Aug. 5, 2026, https://www.geekwire.com/2026/the-startup-idea-that-convinced-a-uw-computer-science-legend-to-leave-google-after-27-years/
- Jeremy Kahn, "Exclusive: Longtime Google DeepMind researcher David Silver leaves to found his own AI startup," Fortune, Jan. 30, 2026, https://fortune.com/2026/01/30/google-deepmind-ai-researcher-david-silver-leaves-to-found-ai-startup-ineffable-intelligence
- Lee Ying Shan and Jennifer Elias, "Google Gemini co-lead Noam Shazeer leaves for OpenAI," CNBC, June 17, 2026, https://www.cnbc.com/2026/06/18/google-gemini-co-lead-noam-shazeer-leaves-for-openai.html
- Bloomberg, "Nobel Prize AI researcher leaves Google to Anthropic," Taipei Times, June 22, 2026, https://www.taipeitimes.com/News/biz/archives/2026/06/22/2003859496
- Amanda Silberling and Lucas Ropek, "AI researchers continue to leave Google for its rivals," TechCrunch, June 24, 2026, https://techcrunch.com/2026/06/24/ai-researchers-continue-to-leave-google-for-its-rivals/
- Reed Albergotti, "DeepMind Chief Demis Hassabis says Google's still winning AI talent," Semafor, June 24, 2026, https://www.semafor.com/article/06/23/2026/deepmind-chief-demis-hassabis-says-googles-still-winning-ai-talent
- Beatrice Nolan, "How stalled models, missed deadlines, and staff burnout led to the unraveling of Google's DeepMind," Fortune, Aug. 10, 2026, https://fortune.com/2026/08/10/how-stalled-models-missed-deadlines-and-staff-burnout-lead-to-the-unraveling-of-googles-deepmind/
- Kai Nicol-Schwarz, "Google's new AI boss inherits a race to catch OpenAI and Anthropic," CNBC, Aug. 12, 2026, https://www.cnbc.com/2026/08/12/google-deepmind-koray-kavukcuoglu.html
- Ali Azhar, "What Google DeepMind's Departures Say About the AI Talent War," AIwire, Aug. 12, 2026, https://www.hpcwire.com/aiwire/2026/08/12/what-google-deepminds-departures-say-about-the-ai-talent-war/