Jeff Dean, Google's chief scientist, is leaving after 27 years. He built the computing and neural-network systems that made Google an early leader in modern AI.
Dean is starting a new venture, Discovery Loop, with three longtime Google colleagues. Google backs it, and it will be a public benefit corporation, which means it aims to help society while making money. He described its mission on X as "to automate machine learning, science, and engineering to accelerate discoveries and progress."
Demis Hassabis is also stepping aside. He founded DeepMind in 2010, and Google bought it four years later. Now he's moving from DeepMind CEO to chairman and taking a new role as Alphabet chief scientist, focused on long-term research and the societal effects of artificial general intelligence, or human-level AI. Koray Kavukcuoglu, DeepMind's tech chief, will run day-to-day work and lead the next Gemini model.
Google published the 2017 paper "Attention Is All You Need," which set off today's generative AI boom. None of those eight researchers remains at the company today.
Noam Shazeer, one of those authors, left for OpenAI in June. Google had spent nearly $3 billion less than two years earlier to bring him back. John Jumper, a Nobel Prize winner, left DeepMind for Anthropic soon after.
Get the free Always Be Buying eBook and learn the simple system for building wealth on any income
Internal approval layers also slow the path from research to product, which makes faster-moving startups more appealing. Gil Luria, an analyst at D.A. Davidson, says the departing researchers want to make history, not commercialize AI. "They're interested in being part of history," he said.
The departures reflect a structural shift: Google has become a frontier model builder, a cloud seller, and a chip supplier to rivals, and those roles increasingly compete for the same resources.
A lot of that squeeze comes down to chips. Google makes its own AI chips, called TPUs, which rival Nvidia's graphics processors. It also sells them to outside labs, including Anthropic, which builds models that compete with Google's Gemini.
So researchers say they can't get enough computing power for their own projects while the cloud unit sells chips to a rival. Assigning TPUs forces choices among model training, product serving, and cloud commitments.
Google spends more on data centers and chips than nearly any other company, but capacity still falls short. The cloud business is booming, and that's the tension.
Thomas Kurian has led Google Cloud since 2019, assembling its commercial engine.
Pichai has said on his two most recent earnings calls that DeepMind's compute needs come first. In July, he called computing power for frontier AI models Google's "first priority" and "the foundation for everything we do." Google is now a $4 trillion company with a lot of places to invest. As Dan Niles of Niles Investment Management puts it, "Somebody's always going to be unhappy in that situation."
Alphabet shares are up 16% this year after rising 65% in 2025, when they beat every other megacap company. They dipped after the earnings report on worries about Google's spending, and again after the leadership news Wednesday. Still, Pichai said 90% of the Fortune 100 use Gemini Enterprise.
The bigger question is whether Google needs to build the best AI models or just sell the tools to whoever does. Tomasz Tunguz, founder of Theory Ventures, says most white-collar work doesn't need the most advanced model: "Many of the models that are reasonable are good enough." Dan Niles agrees: "You don't need a Ferrari for this stuff. A Ford will work for 90% of the use cases."
Google's public line is that it intends to do both. Google can profit by selling infrastructure to labs like OpenAI and Anthropic, selling Gemini to companies, and embedding AI into Search, YouTube, and Workspace.
The question for your portfolio is which Google shows up in the next few years: the one selling the tools, or the one that keeps losing the people who built its edge.
Download the free Always Be Buying eBook and start putting your money to work today