Google panicking? Four top executives jump ship, losing value equivalent to one Jilin Province in a single day.

Author: Bear Cookie, BitpushNews

AI geniuses are collectively defecting from Google.

Noam Shazeer is gone, John Jumper is gone, Jonas Adler and Alexander Pritzel are also gone. In just seven days, Google lost at least four top AI researchers—one went to OpenAI, three went to Anthropic.

The capital market reacted immediately. On June 22, Alphabet, Google's parent company, saw its stock price plummet by 5%, with a single-day market cap loss of approximately $225 billion, equivalent to about 1.53 trillion RMB—this is Google's worst drop in nearly a year, and also the largest single-day market cap loss in the company's history, roughly 1.53 trillion RMB—which equals the total annual GDP of China's Jilin Province.

How can a talent exodus trigger such violent panic in the capital market? Behind this lies the brutal truth of the Silicon Valley AI war.

Who Are the Departing Figures?

Noam Shazeer, Google's Vice President of Engineering and Co-Lead of the Gemini model. You might not be familiar with the name, but his resume is a "living fossil" in the AI world—one of the eight co-authors of the 2017 world-changing paper "Attention Is All You Need." The Transformer architecture, the grandfather of all large language models, he is one of its founders.

Shazeer and Google also share a dramatic "ex-relationship story." In 2021, his repeated efforts to push a chatbot project within Google were blocked, so he quit in frustration and founded Character.AI. In 2022, ChatGPT emerged, and Google was left regretful. In 2024, Google spent about $2.7 billion in a special deal to "invite" him back—Silicon Valley understood that this money bought not technology, but the man himself.

As a result, in less than two years, he left again. OpenAI CEO Sam Altman couldn't hide his satisfaction on X: "Noam is the person I've most wanted to work with since OpenAI's inception. Only waited ten years. I think it's worth it."

John Jumper, Google DeepMind Vice President, core lead of the AlphaFold project, and 2024 Nobel Laureate in Chemistry. Just six months after earning his PhD, Demis Hassabis entrusted him with the AlphaFold mission. Now, the system has predicted over 200 million protein structures, completely rewriting the paradigm of biomedical research.

Jonas Adler, a senior researcher with over 7 years at DeepMind. He participated early in the AlphaFold project, once a teammate alongside Jumper; later shifted to generative AI, becoming a core figure in the development of the Gemini 1.5, 2.5, and 3 series models. Recently, he was mainly responsible for Google's AI coding direction—one of the most critical battlefields for the commercial deployment of large models.

Alexander Pritzel, joined DeepMind in 2014, a veteran of nearly 12 years. He was an early member of DeepMind's core reinforcement learning team, deeply involved in the training architecture of AlphaGo and AlphaZero—those century-defining matches that defeated human Go champions had his code behind them. Later, he brought his reinforcement learning experience to the pretraining field of large language models, the foundational engineering that determines the ceiling of model capabilities.

These four individuals respectively hold the four lifelines of Google AI: Shazeer is the soul of large model underlying architecture, Jumper is the global face of AI for Science, Adler is the key driver of AI coding deployment, and Pritzel is the guardian of model pretraining fundamentals. When these four flags simultaneously fell to direct competitors within a week, the capital market could hardly stay calm.

An even more cutting judgment is: These people may not have left for money, but rather because they believe they have a better chance of being at the forefront of research at OpenAI or Anthropic than at Google.

How Normal Is AI Talent Hopping in Silicon Valley?

This is almost the underlying operating logic of Silicon Valley.

Unlike strict non-compete agreements in China, Silicon Valley has long adhered to the principle of free mobility.

From the "Traitorous Eight" of Fairchild Semiconductor founding Intel, to Facebook poaching Sheryl Sandberg from Google, talent migration has always been the core driving force of Silicon Valley innovation.

During the dot-com bubble of the late 1990s, countless engineers flocked from IBM, Microsoft, and Oracle to startups, giving birth to giants like Google, Amazon, and eBay. As early as 2011, statistics showed that companies suffering the most severe talent loss included Yahoo and MySpace, while the most aggressive poachers were Facebook and Twitter.

Today's AI talent war follows the same script—different technological waves, the same underlying logic: top talent always flows to where they believe they can define the next era.

Moreover, OpenAI has already submitted a confidential IPO application to the SEC, and Anthropic is also in line. Employees of both companies hold substantial equity. Shazeer and Jumper chose this window to jump ship, timing it perfectly.

According to incomplete statistics, from January 2025 to now, at least 32 core researchers above the director level have changed hands among Silicon Valley AI giants. When Meta established its superintelligence lab in July 2025, it poached 11 people at once from OpenAI, Anthropic, and Google. Talent mobility has long been routine.

The only difference is that current compensation can no longer be described as "high salary." According to Bloomberg, Meta once offered a signing bonus as high as $100 million to poach a core researcher from OpenAI, almost a sports league-level "transfer fee."

This is the "celebrity era" of the AI talent war—the giants are no longer competing for ordinary researchers, but for "superstars" who can change technical directions and attract team and capital confidence.

Is Google's Moat Still Deep Enough?

But can a single-day market cap evaporation of this magnitude really be caused by just a few people leaving?

Of course not entirely. Behind it is collective panic triggered by multiple negative factors stacking up.

On the same day, Alphabet's Gmail and YouTube experienced large-scale access failures, further dampening investor confidence. Microsoft CEO Satya Nadella's recent remarks about the "commoditization" of the AI market also made the market reassess the return on Google's massive AI spending.

There is also a signal overlooked by most—the shadow of an AI price war is approaching.

Bill Maris, founder of Google Ventures, publicly stated on the All-in podcast: "If I were Google and decided to arbitrarily cut token prices by 80%, what would happen to OpenAI's and Anthropic's business models?" This remark directly hit Wall Street's most sensitive nerve: Google has over $300 billion in annual advertising revenue and can easily subsidize its AI business with core profits. If Google truly launches a price war, how much profit margin would be left for the entire industry?

Large model companies are valued near one trillion, but so far no one has delivered a convincing commercial report card. C-end user willingness to pay is extremely limited—if Doubao charges, users immediately switch to DeepSeek or Kimi; B-end clients are penny-pinching, and Alphabet itself admits that if enterprise clients migrate 80% of their workloads to Gemini, they can save $1 billion annually.

This is not an isolated case but an industry-wide "prisoner's dilemma." Both OpenAI and Anthropic, while facing high computing costs, once drawn into homogenized low-price competition like "delivery riders," how to maintain a valuation logic as high as trillions of dollars has become an unavoidable business question.

During the dot-com bubble, the survivors were not the most hyped portals, but those that built ecosystems: Google, Facebook, Amazon, Alibaba, Tencent.

In today's AI industry, who will have the last laugh? The answer is not yet clear. But one thing is certain: first grab the smartest people, then talk about the rest.

This AI talent war is far from over.