谷歌病了

Original author: Sleepy

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

Shazeer is a core author of the landmark 2017 paper "Attention Is All You Need" and 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, staying for over two decades. He left in 2021 to start his own company after Google refused to release his chatbot, Meena.

Google paid a huge sum to bring him back, giving him the title of Vice President of Engineering and making him a co-leader of Gemini, hoping he could help Google win the AI war.

Less than two years later, he left. He went to OpenAI.

According to Bloomberg, shortly before Shazeer announced his departure, the computing resources for one of his projects were internally reallocated by Google to the DeepMind team. Insiders said this adjustment was aimed at promoting team collaboration and consolidating pre-training work.

The Nobel Laureate's Farewell

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

Jumper's story is different from Shazeer's. Shazeer was a veteran, having spent over twenty years at Google, witnessing all the company's highs and lows. But Jumper was nurtured by the place.

Just six months after Jumper earned his PhD, Hassabis made a risky decision, entrusting the young, inexperienced scientist to lead the entire protein structure prediction project.

Jumper didn't waste the opportunity. He led his team to create 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 about trust and fulfillment. Hassabis trusted a young man, and the 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 going to Anthropic.

When the market opened on 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 anxiety over the AI competition. These two departure announcements were the final straw.

Over the next few days, news came one after another. Jonas Adler and Alexander Pritzel were also leaving, heading to Anthropic. These two were core contributors to Gemini and also former partners of Jumper on the AlphaFold project. Adding AI safety researcher Arthur Conmy, who had left earlier, Google lost at least five top researchers in one month, four of whom went to Anthropic.

Hassabis had nurtured Jumper from the start, and now he watched him walk into a competitor's door with half of the AlphaFold team. I don't know what he saw under Jumper's tweet, but I can guess it probably felt like a familiar sense of fate.

Nursery

The best technology companies of each generation eventually become the nurseries for the next.

Google itself grew up this way.

Many of its earliest engineers came from Microsoft, IBM, Yahoo, and Bell Labs. When Microsoft was battered by antitrust cases in the 2000s, a large number of top talents flowed to Mountain View, including a young Noam 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 to every corner of Silicon Valley, becoming the founding teams of others' companies.

Now it's Google's turn.

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

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

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

Back then, no one asked, "Can Google win the AI war?" because asking that was as superfluous as asking, "Will the sun rise in the east?" Google had the best researchers, the most data, the most powerful computing, and the most money. If it wasn't going to win, who was?

But look at who is now standing opposite Google.

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

Anthropic's founders, the Amodei siblings, previously did safety research at OpenAI, and OpenAI's early core team itself included many people from Google.

Jumper spent nine years at DeepMind. Shazeer spent over two decades at Google. Tracing the talent chain of the entire AI industry to its source, almost everyone has worked in Mountain View at some point.

In 2025, SignalFire conducted a study showing that DeepMind engineers were 11 times more likely to jump to Anthropic than in the reverse direction.

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

Google provides the money, computing power, and free environment, recruiting the brightest young people in the world, giving them the best conditions to do cutting-edge research. Once they have their wings, they fly away, go to the other side, build better products than yours, and come back to compete against you.

Can't Keep the Doers

Google's problem isn't just retaining talent. It retained Shazeer when it spent $2.7 billion to buy him back. The problem is what happened after it kept him.

Shazeer left Google twice.

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

Both departures were essentially because he wanted to build things, but the organization wouldn't let him.

Google's decision-making chain is too long. For a new AI feature to go from R&D to launch, it must pass through product, legal, compliance, PR, and business unit approvals. Getting stuck at any level means months of delay. By the time technology developed in the DeepMind lab finally enters a consumer product, the window of opportunity has already passed.

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

But merging isn't the same as integrating. The two teams' respective codebases, data flows, and work habits haven't been fully unified even today. The reallocation of Shazeer's computing resources to the DeepMind team is a microcosm of this "merged but not integrated" state. Nominally one department, but how resources are divided and who sets priorities remains an internal game.

When an organization can't effectively utilize talent, its products naturally get worse. Google Search's AI Overviews feature once suggested users put glue on pizza to prevent cheese from sliding off, said running with scissors is a good cardio workout, and confidently answered "No, it's 2025" when asked "Is it 2026?". Research showed it was generating tens of millions of incorrect answers per hour.

In early 2025, Google announced it would fully migrate Google Assistant to Gemini. Basic features that had worked for nearly a decade suddenly malfunctioned – setting alarms and controlling smart homes all had issues. The migration, originally scheduled for completion by the end of the year, had to be pushed to 2026.

In July of the same year, Google's newly launched Gemini CLI coding tool caused an accident. A user asked it to organize a folder, and it hallucinated a series of non-existent operations, deleting all the project files. Afterwards, it admitted, "I have completely and catastrophically failed you."

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

None of these are deep technical problems. Permission isolation, feature regression testing – a properly functioning engineering team wouldn't stumble on these things.

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

But I think it's too glib to simply attribute this problem to "systemic issues."

The very system that allowed Jumper to spend nine years perfecting AlphaFold is Google's system. Don't push for commercialization, don't cut the budget, don't ask when results will come. This kind of patience and 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 working on something with no guarantee of success. AlphaFold could never have been born in a place with weekly iterations.

But the problem is that this same depth, while protecting you to do AlphaFold, also accumulates layers of approval, departmental interests, and compliance processes. While giving you nine years of freedom, it also grows those twelve layers of political maneuvering that prevent you from getting computing resources.

The soil that nurtures genius is the same soil that traps it. This is almost inevitable for an organization that has grown to this scale and achieved this level of success.

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 isn't good to them, but because they have become the very thing they least wanted to be at Google: capable and ambitious people who cannot get things done.

But who knows, maybe twenty years from now, some young person at Anthropic will also post a tweet saying they are leaving to join a company founded just three years ago.

If You Don't Go Out and See the World

On June 23rd, Hassabis was interviewed at the Cannes Lions Festival and was asked about the recent talent exodus.

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

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 has lost and better than anyone why he couldn't keep them. So I don't think he was just putting on a brave face when he said this. Maybe it's a man who sees the ending, giving himself a final moment of dignity.

It reminds me of the line from Cinema Paradiso, where the old projectionist Alfredo says to the young Toto:

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

When Alfredo said this, he was pushing Toto away. No one wanted to keep the child more than he did, 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 light and shadow, his initial curiosity about the world. But that's all the cinema could give. For the rest of the journey, he had to go out and find it.

Google was once the cinema paradiso for all AI researchers. The best equipment, the most relaxed environment, the most knowledgeable colleagues. You could spend nine years building a model to predict protein structures without rushing to commercialize it, without writing PowerPoint presentations 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 for you. Everyone thought Google was the whole world back then.

But perhaps a place's finest hour is precisely the time it should send its people away.

Today, the free cafeterias in Mountain View still serve three meals a day. The colorful bicycles are still parked at the entrance of every building, ready for anyone to ride. Every week, a new cohort of Nooglers puts on the iconic propeller hats and takes a group photo, their eyes bright and full of promise.

Just like Shazeer walking into Google for the first time twenty years ago, and Jumper joining 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