Two resignations, a Transformer co-author bound for OpenAI and a Nobel laureate bound for Anthropic, triggered Alphabet's worst session in over a year and put its costly AI bet under the microscope.
Alphabet shareholders got an expensive reminder on Monday about the scarcest asset in the AI race: the small number of people who can build the next breakthrough.
The Google parent shed somewhere between roughly $250 billion and $270 billion in market value on June 22, depending on which intraday and closing figures you use, in its steepest single-day drop since May 2025. Closing estimates cluster around a 5% to 7% decline. The stock had touched an all-time high of $408.61 on May 18 before sliding to about $362 by Monday morning, leaving it close to 11% below that peak.
The trigger was not an earnings miss or a product flop. It was two resignation posts on X.
The two posts that rattled Wall Street
The first came from Noam Shazeer, a vice president of engineering and co-lead of Google's Gemini models, who said on June 18 that he was leaving for OpenAI. The defection stings for reasons that go well past one salary. Shazeer co-authored the 2017 paper "Attention Is All You Need," the research that introduced the Transformer architecture now sitting under almost every large language model in production, Gemini and ChatGPT among them. Google had also paid dearly to get him back. In September 2024 the company struck a deal worth about $2.7 billion for Character.AI's technology, an arrangement Silicon Valley widely read as a way to rehire Shazeer and his team under the cover of a licensing agreement. He stayed less than two years.
A day later, the second post landed.
John Jumper, a DeepMind vice president and engineering fellow, announced he was leaving after nearly nine years to join Anthropic, planning to recharge before he starts. Jumper is no routine hire. He shared the 2024 Nobel Prize in Chemistry with DeepMind chief Demis Hassabis for AlphaFold, the system that has predicted more than 200 million protein structures and changed how biologists approach their work. For Anthropic, signing the researcher behind AI's most celebrated scientific result reads as a statement about where it intends to compete.
The two exits did not happen in isolation. Andrej Karpathy, an OpenAI co-founder, joined Anthropic in May. Stacked together, the moves sketch a pattern that has unsettled investors: senior researchers picking funded challengers over the AI divisions of trillion-dollar incumbents, sometimes over challengers their former employer helps bankroll. Google, for its part, still holds a sizable stake in Anthropic.
Why researchers now move markets
Why should two departures erase a quarter of a trillion dollars from a company worth more than $4 trillion? Because the way the market values AI leaders has quietly changed.
For decades, technology firms were priced on products, patents, and market share. The AI buildout has added a fourth input that is far harder to put on a balance sheet: the handful of people who can design the next model architecture. Chris Beauchamp, chief market analyst at IG, described figures like Jumper and Shazeer as "pearls of great price," arguing their exits ripple far beyond their own teams. Gil Luria of D.A. Davidson put the competitive read more bluntly, telling Barron's that "the race at the frontier right now appears between Anthropic and OpenAI." His broader point was about scarcity. There is so much demand for elite researchers that frontier labs will pay almost anything to land one.
Compensation has started to look like professional sports. OpenAI's Sam Altman has said Meta dangled signing bonuses as high as $100 million to pull away his staff. Anthropic, valued near $1 trillion in recent funding talks and the subject of heavy IPO speculation for late 2026, has leaned on a different pitch, betting that mission and retention beat raw cash.
The talent scare collided with a worry that was already nagging at investors: how much Google is spending to stay in front. Alphabet lifted its 2026 capital expenditure guidance to a range of $180 billion to $190 billion. For scale, that is roughly six times the $31 billion it spent in 2022 and about double the $91 billion it laid out in 2025. To fund the buildout, the company announced an equity raise in early June that started at $80 billion and was upsized within two days to nearly $85 billion as demand ran hot, with $10 billion of it coming from Warren Buffett's Berkshire Hathaway. Melius Research has cautioned that Alphabet's free cash flow could turn negative for the next few years as the spending ramps. The uncomfortable question writes itself. If rivals are simultaneously poaching the researchers meant to justify that outlay, what exactly is the money buying?
Hanging over all of it was a warning from a competitor. In a Wall Street Journal interview published the day before the selloff, Microsoft chief Satya Nadella argued that the AI market is sliding toward commoditization and cautioned against the industry's growing dependence on a handful of dominant labs. He has been sketching an alternative in which customers mix lower-cost models from different providers rather than chaining themselves to one lab, a vision Microsoft is backing with products like its Copilot Cowork agent and reported interest in hosting cheaper models from China's DeepSeek. If frontier models keep getting cheaper and more interchangeable, the logic runs, then heavy spending stops buying a lasting lead. Nadella also pushed back on the darker talk about AI erasing white-collar work, suggesting companies redesign roles instead of cutting them, and said the sector needs what he called "social permission" to keep expanding.
What Google still has going for it
None of this means Google's AI program is unraveling, and plenty of analysts said so.
The company's most recent quarter was strong by almost any measure. Revenue hit $109.9 billion, up 22% from a year earlier, while operating income climbed 30% to $40 billion and extended a run of eleven straight quarters of double-digit growth. Google Cloud grew 63% to roughly $20 billion. Gemini's consumer app crossed 750 million monthly users late last year and was handling more than 16 billion tokens a minute by the first quarter. Analysts still lean heavily bullish, with buy ratings far outnumbering sells and an average target north of $430. When Shazeer's move first broke, the stock actually closed higher that session. The panic only set in once Jumper followed and the storyline shifted from one resignation to a trend. CNBC's Jim Cramer had flagged the Shazeer hire early, calling it "a coup" for OpenAI and noting how oddly muted Alphabet's reaction was at the time.
DeepMind also remains one of the deepest research labs anywhere, and Alphabet has something its private rivals still envy: search and YouTube distribution measured in billions of users, alongside profitable businesses that can bankroll years of infrastructure without waiting on the next funding round.
The legal backdrop is getting more crowded too. The EU's AI Act moves into its most demanding phase for high-risk systems in August, with penalties that can reach 7% of global revenue, while a UK transparency order on Google Search and pending US ad-tech remedy hearings add to the load. A fresh $75 million DeepMind partnership with film studio A24 did little to brighten the mood on Monday.
The cleanest signal to watch from here is Google's release cadence. If Gemini updates keep shipping on schedule through the back half of 2026, the market will treat these departures as noise around a still-dominant program. If the pace slips, investors will start asking whether the people who walked out were carrying more of the roadmap than a $4 trillion balance sheet can quickly replace.
Comments
Join the discussion and share your perspective.