Google is hawking its most powerful AI chips to outside bidders—including direct competitors—while the researchers who built the company's AI legacy can't get enough compute to do their jobs, and they're walking out the door. The most powerful technology since the printing press is being allocated by a handful of trillion-dollar companies, and the American public has zero say in who wields it.
Google Cloud just posted 82 percent second-quarter revenue growth, outpacing Amazon and Microsoft. The engine driving that growth: selling Google's proprietary TPU chips to outside AI labs, including Anthropic, whose models compete directly with Google's own Gemini. Meanwhile, Google's internal researchers say they can't get the chips they need for ambitious projects. When the company announces massive infrastructure commitments to rivals, frustration inside the labs runs especially hot.
The people who made Google an AI pioneer have had enough. Chief Scientist Jeff Dean is departing after 27 years, taking senior engineers Sanjay Ghemawat, Oriol Vinyals, and Quoc Le to found Discovery Loop, a Google-backed public benefit corporation aimed at automating machine learning and scientific discovery. With them goes the last of the eight authors of Google's landmark 2017 transformer paper—the work that launched the entire generative-AI era. Every single one has now left.
Demis Hassabis, who co-founded DeepMind and sold it to Google in 2014, stepped down as Google DeepMind CEO to become the unit's chairman and Alphabet's chief scientist, focusing on longer-term research and the societal implications of artificial general intelligence.
The departures aren't just about compute access. Breitbart News reported that Noam Shazeer, another founding engineer of the AI revolution and co-author of the transformer architecture papers, previously left for OpenAI after clashing with Google's leftist workforce over his personal beliefs. Shazeer wrote in internal posts that he doesn't believe humans "have an attribute called gender" and that "it is not okay to sterilize children." Such common-sense opinions, Breitbart noted, are not typically found within Google—a company that fired engineer James Damore for his political views. The message to independent-minded researchers is clear: conform or leave.
Bureaucracy compounds the frustration. Multiple layers of approval are needed to move research into products, making nimbler outfits like OpenAI and Anthropic more appealing to scientists who want to build rather than navigate corporate politics.
Gil Luria, a D.A. Davidson analyst, said the pattern is obvious: "They're not interested in commercializing AI. They're interested in being part of history, and so they look at Anthropic, OpenAI or another startup as being the place where they can pursue history."
Google shareholder Dan Niles framed the business case bluntly: "The models are good enough for 90% of what needs to get done. You don't need a Ferrari for this stuff. A Ford will work." Selling the compute is a surer bet than chasing the frontier yourself.
Alphabet CEO Sundar Pichai insists securing frontier compute is the company's "first priority." Both things can be true—Google can call the frontier a priority and still find that when cloud revenue is this lucrative, the researcher chasing the next breakthrough waits in line behind the customer writing the check.
TNW framed the story as a strategic dilemma—whether Google should build the best model or just sell the infrastructure. Breitbart added what TNW buried: Google's woke internal culture is itself a driver of the exodus, pushing out talent that won't toe the ideological line. The real story is both. A trillion-dollar company is choosing monopoly rents over innovation, and a rigid ideological orthodoxy is accelerating the brain drain that innovation requires.
The question isn't whether Google can keep its cloud profits rising. It's who controls the infrastructure that will shape the next century—and whether Americans will have any voice in that answer before it's settled without them.
