Google Lost 4 Top AI Researchers In 6 Days, And Anthropic Took Three Of Them

Welcome back to SavvyMonk, your one-stop for AI and tech news that actually matters.

Google has spent years as the place the best AI researchers wanted to work, and in the space of 6 days in June 2026 four of them decided otherwise. Three went to Anthropic, one went to OpenAI, and the bill arrived as a stock drop worth roughly a quarter of a trillion dollars.

Let's get into it.

Managing your cap table doesn’t have to be complex. Pulley simplifies equity management for Founders and CFOs with intuitive workflows, accurate, audit-ready reporting, and predictable pricing—so you can plan and scale without surprises. Onboard in days, not weeks, and rely on responsive, expert support every step of the way.  

From issuing grants to 409A valuations or ASC 718 reporting, Pulley gives you the clarity to manage equity, make decisions, and get back to work. Experience a platform built for business owners and finance teams: transparent, reliable, and designed to put your company first.

See Pulley in action

The exits came in a tight sequence that made the pattern impossible to miss. On 18 June 2026, Noam Shazeer announced he was leaving Google for OpenAI. One day later, on 19 June, John Jumper said he was leaving Google DeepMind for Anthropic after nearly nine years.

Then on 24 June, two more Gemini contributors, Jonas Adler and Alexander Pritzel, were reported to be preparing to follow Jumper to Anthropic. The two have not announced the move themselves the way Jumper and Shazeer did, though the reporting fits a pattern that is now hard to read as coincidence. What ties all four together is that they came from the teams behind Gemini, the model Google treats as its most important.

Who Walked Out

These are not interchangeable engineers. Shazeer was a vice president of engineering and a co-lead of Gemini, and he co-wrote the 2017 paper that introduced the transformer architecture now sitting underneath nearly every modern large language model in production.

Jumper shared the 2024 Nobel Prize in Chemistry with DeepMind chief Demis Hassabis for AlphaFold, the protein structure system that has predicted more than 200 million structures and reshaped how drug discovery and structural biology get done.

Adler worked on Google's AI coding effort, the area most directly relevant to competing with tools like Cursor and Claude Code, while Pritzel worked on pretraining, the early stage where a model learns from large volumes of data.

Losing one of these people in a quarter would register, and losing all four inside a week reads less like coincidence and more like a verdict.

The 2.7 Billion Dollar Boomerang

Shazeer's exit carries the most awkward history, because of what Google paid to get him back in the first place. He left the company in 2021 after it declined to launch a chatbot he had built, then co-founded Character.AI.

In 2024 Google brought him back through a 2.7 billion dollar licensing deal that worked in practice as a reacquisition, and it installed him as a co-lead on Gemini. Less than two years later he is at OpenAI, where chief executive Sam Altman welcomed him publicly as one of the people he had most wanted to work with.

A company that spends 2.7 billion dollars to reverse a single departure, only to watch that person leave again for its closest rival, has a retention problem that money alone has not solved.

What It Cost Alphabet

The market reaction was fast and large. On Monday 22 June, Alphabet stock fell as much as 7.2 percent during the day and closed down around 6 percent, its worst trading day in about a year. Estimates of the market value erased run from roughly 225 billion dollars to 269 billion dollars depending on how the moment is measured, which is a striking price to put on two resignations.

The slide landed in a jittery week, following a weekend interview in which Microsoft chief Satya Nadella argued the AI market was becoming commoditised, and it arrived while Alphabet was already under scrutiny for raising 141 billion dollars in debt and equity since October and projecting between 180 and 190 billion dollars of AI capital spending for 2026. D.A. Davidson analyst Gil Luria framed it plainly, warning that Google is losing the war for talent at the frontier and may now be falling visibly behind.

Why They Keep Leaving

The pull is not only salary, and the structural reason is worth understanding. Both Anthropic and OpenAI are close to going public, which lets a researcher join before an initial public offering and own a slice of the upside, something a company already worth around 2 trillion dollars cannot easily match.

Anthropic raised 65 billion dollars on 28 May 2026 at a post-money valuation near 965 billion dollars and has been reported to be in talks at close to 1 trillion dollars, so the pre-IPO equity on offer is real rather than theoretical. Compute access matters as well, and the timing tells its own story, since shortly before Shazeer announced his move Google reportedly reassigned computing capacity from one of his projects to another DeepMind team in London.

A researcher who cannot get the hardware to run experiments has a reason to take the call when a rival comes knocking.

What Anthropic Is Actually Buying

Reading these hires as scattered poaching misses what is happening, because the targeting is specific. Adler's coding work maps directly onto Anthropic's push behind Claude Code against the wider field of AI coding tools, Pritzel's pretraining expertise sits at the core of building a frontier model, and Jumper's protein work fits Anthropic's expansion beyond general chatbots into healthcare and scientific applications.

Taken together, Anthropic is acquiring institutional knowledge of how Gemini was built rather than four loose résumés, and that knowledge is the part Google cannot simply backfill with a job posting. The startup has also been recruiting from DeepMind at a steady clip, and industry data cited in the reporting suggests researchers move from Google to Anthropic far more often than the reverse.

The Part Worth Keeping in Proportion

None of this unmakes Google, and the honest read needs the counterweight. Google DeepMind still employs thousands of researchers across Search, Android, YouTube and Cloud, Gemini remains a genuinely competitive model, and Alphabet's profitable businesses can fund years of infrastructure without leaning on outside capital rounds the way its rivals do.

Jumper is reportedly not due to start at Anthropic until 2027, two of the four moves are still reported rather than confirmed, and analysts at Jefferies waved the week off as noise amid intense competition. The fair summary is that four departures do not break a research operation this deep, while they do hand the next researcher weighing an offer a fresh reason to wonder where the most interesting work is now happening.

The Bottom Line

Four exits in 6 days does not erase Gemini, and Google has both the cash and the history to refill the seats. The trouble is that the people leaving are the ones who know how the flagship was built, and they are walking to the two rivals best placed to use that knowledge.

The 269 billion dollar drop is the market pricing a single uncomfortable question, which is whether the next researcher deciding whether to stay still believes the frontier runs through Google.

Category: Career Decision

"I have an offer from [company] and a competing offer or current role at [company]. Here are the two packages [paste salary, equity, vesting schedule, and any pre-IPO terms]. Walk me through the real value of each, treating the private company equity as a range rather than a single number, list the three biggest risks hiding in the pre-IPO offer, name the one question I should ask about compute or resources before I decide, and tell me which offer wins if I care most about [impact, money, or stability]. Give me one short answer for each."

The headline number is the 269 billion dollars, and the quieter story underneath it is that AI leadership is now being priced on people rather than products. A model can be matched and a chip can be bought, but a researcher who knows how the hardest problems were solved cannot be replaced with a software update.

Google can backfill the seats quickly enough, and the harder task is convincing the people still sitting in them that the best work is happening where they already are. Hit reply, I read every response.

See you in the next one.

— Vivek

P.S. Know someone who follows the AI talent wars or is weighing an offer from a frontier lab? Forward this to the friend who reads every funding announcement before breakfast. They can subscribe at https://savvymonk.beehiiv.com/