谷歌กำลังป่วย

Original author: Sleepy

In August 2024, Google spent $2.7 billion to bring Noam Shazeer back from Character.AI, the company he founded.

Shazeer was a core author of the 2017 paper "Attention Is All You Need," a co-inventor of the Transformer architecture. Without his paper, there would be no GPT, no Claude, no Gemini, and no AI industry as we know it today.

He joined Google in 2000 as one of its earliest employees and stayed for over two decades. He left in 2021 to start his own venture after Google refused to release his chatbot, Meena.

Google paid a high price to bring him back, giving him the title of Vice President of Engineering and appointing him as co-leader of Gemini, hoping he could help Google win the AI race.

Less than two years later, he left. To OpenAI.

According to Bloomberg, shortly before Shazeer announced his departure, computing resources for one of his projects were reallocated internally by Google to the DeepMind team. Insiders say this adjustment was intended to promote team collaboration and integrate pre-training efforts.

The Nobel Laureate's Farewell

Shazeer left on June 18. The next day, John Jumper also left.

Jumper's story is different from Shazeer's. Shazeer was a veteran, having spent over two decades at Google, witnessing both the good and the bad. But Jumper was nurtured by this place.

Just six months after Jumper received his PhD, Hassabis made a risky decision, entrusting this young man with no management experience to lead the entire protein structure prediction project.

Jumper did not let this opportunity slip. He led his team to develop AlphaFold, predicting the three-dimensional structures of over 200 million proteins, advancing the entire field of structural biology by a decade. In 2024, he stood alongside Hassabis in Stockholm to receive the Nobel Prize in Chemistry.

The first half of this story is one of trust and fulfillment. Hassabis trusted a young man, and that young man repaid him over nine years, repaying all of human biology. But there's a second half. Two years after winning the Nobel Prize, on June 19, 2026, Jumper posted a short tweet saying he was leaving for Anthropic.

When the market opened the following Monday, Alphabet's stock price plummeted. It fell as much as 7% during the day, closing down about 5%, wiping out approximately $225 billion in market capitalization – losing the value of a Spotify. Alphabet's stock had been declining since hitting an all-time high in early 2026, weighed down for months by antitrust lawsuits, massive capital expenditures, and anxieties over the AI competition. These two departure announcements were the final straw.

Over the next few days, announcements came one after another. Jonas Adler and Alexander Pritzel were also leaving, heading to Anthropic. They were core contributors to Gemini and long-time partners of Jumper on the AlphaFold project. Adding to these losses was AI safety researcher Arthur Conmy, who had left earlier. Within a month, Google lost at least five top-tier researchers, four of whom joined Anthropic.

Hassabis once nurtured Jumper from the ground up. Now he watches him lead half the AlphaFold team into a competitor's camp. I don't know what he saw under Jumper's tweet, but I can guess it was a familiar sense of destiny repeating itself.

Nursery

Every generation's best tech company inevitably becomes the nursery for the next generation.

Google grew up this way itself.

Many of its earliest engineers came from Microsoft, IBM, Yahoo, and Bell Labs. In the early 2000s, when Microsoft was weakened by antitrust battles, a large number of top talents flowed to Mountain View, including a young Shazeer.

Going further back, Bell Labs invented the transistor, Unix, and the C language, laying the foundation for the entire information age. But what about Bell Labs itself? Its people scattered across every corner of Silicon Valley, becoming founding teams of other companies.

Now it's Google's turn.

In 2016, AlphaGo defeated Lee Sedol, and the world learned for the first time that AI could achieve such feats. That was Google's moment.

In 2017, the Transformer paper was published, laying the foundation for the entire AI industry. That was still Google's moment.

In 2021, AlphaFold predicted 98% of human protein structures. That, too, was Google's moment.

Back then, no one asked, "Can Google win the AI race?" Because asking that was as redundant as asking, "Will the sun rise in the east?" Google had the best researchers, the most data, the strongest computing power, and the most money. If it didn't win, who would?

But look now. Who is standing opposite Google?

OpenAI co-founder Ilya Sutskever, who did deep learning research with Geoffrey Hinton at Google in his early years.

Anthropic founders Dario and Daniela Amodei, who previously worked on safety research at OpenAI. And OpenAI's early core team itself included many ex-Google people.

Jumper spent nine years at DeepMind; Shazeer spent over two decades at Google. The entire talent chain in the AI industry, traced to its source, almost all once worked in Mountain View.

A 2025 study by SignalFire found that DeepMind engineers were 11 times more likely to jump ship to Anthropic than the reverse.

Someone commented on this wave of departures on Twitter, writing: "Google is turning into Anthropic's training ground."

Google provides the funding, computing power, and free environment, recruiting the world's brightest young people, giving them the best conditions for cutting-edge research. When they grow strong enough, they fly away, go to the other side, build better products than yours, and come back to compete.

Can't Keep Doers

Google's problem isn't just retaining talent. At the moment it bought back Shazeer for $2.7 billion, it did retain him. The question is, what happened after that?

Shazeer left Google twice.

First in 2021, when Google refused to release his chatbot Meena. ChatGPT hadn't been born yet, and Google was cautiously observing conversational AI. Shazeer couldn't wait, so he left on his own. The second time is now. His computing resources were reallocated, and he left again.

Both departures stemmed from the same issue: he wanted to build things, but the organization wouldn't let him.

Google's decision-making chain is too long. A new AI feature, from development to launch, must pass product, legal, compliance, PR, and business line approvals. Getting stuck at any level means months of delay. By the time technologies developed in DeepMind labs finally enter consumer products, the window of opportunity has often passed.

In 2023, Google merged DeepMind and Google Brain, and everyone was optimistic about this combination of the two strongest AI teams.

But merger is not integration. The two teams' respective codebases, data flows, and work habits remain unbridged to this day. Shazeer's computing resources being reallocated to the DeepMind team is a microcosm of this 'merged but not integrated' state. Nominally one department, but resource allocation and priority setting are still internal battles.

When an organization can't effectively utilize talent, its products naturally deteriorate. Google Search's AI Overviews once suggested users put glue on pizza to prevent cheese from sliding off, stated that running with scissors is a cardiovascular exercise, and confidently answered "No, it's 2025" when asked "Is it 2026?" Research showed it generated tens of millions of erroneous answers per hour.

In early 2025, Google announced the full migration of Google Assistant to Gemini. Basic features that had worked for nearly a decade suddenly malfunctioned – setting alarms, controlling smart home devices all had issues. The migration, originally scheduled for completion by the end of the year, had to be postponed to 2026.

In July of the same year, Google's newly launched Gemini CLI coding tool caused another incident. A user asked it to organize a folder, and it hallucinated a series of nonexistent operations, deleting all the project files before confessing, "I have completely and catastrophically failed you."

By the I/O conference in May 2026, CEO Sundar Pichai confidently stated that Gemini 3.5 Pro would launch "next month," but it was later postponed to July.

None of these are deep technical problems. Permission isolation, functional regression testing – a well-functioning engineering team wouldn't stumble on these issues.

Bad products and talent exodus are two sides of the same coin. The organization can no longer channel the impulses of genius into products. The technology and people are still there, but the path from idea to launch is blocked.

But attributing this solely to "systemic issues" feels flippant.

It was precisely Google's system that allowed Jumper to spend nine years perfecting AlphaFold. No pressure to commercialize, no budget cuts, no demands for immediate results. This patience, this depth, is something no startup can provide.

Anthropic and OpenAI can let you iterate every two weeks, but they can't let you spend nine years on something with an uncertain outcome. AlphaFold could not have been born in a place with a weekly iteration cycle.

But here's the paradox: this same depth, while protecting your ability to create AlphaFold, simultaneously builds up layers of approval, departmental interests, and compliance procedures. While giving you nine years of freedom, it also fosters the intricate politics that prevent you from getting computing resources.

The soil that nurtures genius is the same soil that traps it. This is an almost unavoidable fate for an organization that has grown to this size and won to this extent.

What Anthropic and OpenAI offer is precisely a place where an idea can directly become action, plus pre-IPO equity. People leave not because Google treated them poorly, but because at Google, they have become the very type of person they never wanted to be: capable and ambitious, yet unable to execute.

But who knows? Maybe twenty years from now, some young person at Anthropic will post a tweet saying they're leaving, heading to a company that's only been around for three years.

If You Don't Go Out

On June 23, Hassabis was interviewed at the Cannes Lions International Festival of Creativity and was asked about the recent brain drain.

He said: "Talent flow between major labs is normal. We have our share of top-tier talent. We possess the largest and most diverse research team among all labs."

Hassabis is one of the smartest people in this industry. He personally guided Jumper from a fresh PhD graduate to a Nobel laureate. He knows better than anyone what he is losing, and better than anyone why it's impossible to retain them. So, I think he wasn't just being stubborn when he said that. Perhaps it was a person who sees the endings trying to preserve a final shred of dignity.

I'm reminded of the film *Cinema Paradiso*, where the old projectionist Alfredo says to young Toto:

"If you don't go out and see the world, you'll think this is all there is."

Alfredo said this while pushing Toto away. He was the one who cherished the boy most, but he knew that staying in that small-town cinema, Toto would never become the person he was meant to be. The cinema gave Toto everything: his love for film, his understanding of imagery, his initial curiosity about the world. But that was all the cinema could offer. The rest of the journey had to be walked out beyond its walls.

Google was once that Cinema Paradiso for every AI researcher. The best equipment, the most permissive environment, the most knowledgeable colleagues. You could spend nine years building a model to predict protein structures without rushing to commercialize it or writing PPTs for executives. When you succeeded, the entire field of biology would applaud you. You would accept an award in Stockholm, and the whole company would celebrate. Everyone thought Google was the whole world.

But perhaps the best time for a place *is* the time it should send its people away.

Today, the free cafeteria in Mountain View still serves three meals a day. The colorful bikes are still parked outside every building for anyone to ride. Every week, a new cohort of Nooglers puts on the iconic propeller hat and takes a group photo, eyes bright.

Just like Shazeer, who walked into Google for the first time twenty years ago. Just like Jumper, who joined DeepMind nine years ago.

References

[1] Google Paid $2.7 Billion to Bring Back an AI Genius Who Quit in Frustration, The Wall Street Journal

[2] Attention is All You Need, Google Research

[3] Top AI researcher leaves Google for OpenAI, Axios

[4] After nearly 9 years, I have decided to leave Google DeepMind and join Anthropic, John Jumper/X

[5] Google poised to lose two more high-profile AI staffers to Anthropic, Los Angeles Times/Bloomberg

[6] AI researchers continue to leave Google for its rivals, TechCrunch

[7] Alphabet sees $225 billion market-cap wipeout as investors fear it's losing the war for AI talent, MarketWatch

[8] Some Reasons Why Google Had Such A Bad Day, The Wall Street Journal

[9] Google's Brain Drain Deepens: Alphabet Braces for Second Day of Losses on Anthropic Poach, Barron's

[10] AI lab musical chairs hits Google the hardest, Axios