Inside the AI compute crunch driving Google researchers to quit
Google's AI researchers are facing challenges in accessing the computing power they need due to high demand from both internal projects and external customers. This scarcity has led some top researchers to leave the company for startups that offer more resources and fewer bureaucratic obstacles. As a result, the competition for Google's tensor-processing units (TPUs) is influencing research directions and project priorities within the company.
- ▪Google's in-house chips and cloud business have created a competitive environment for computing resources.
- ▪AI researchers at Google are reportedly losing access to computing power in favor of revenue-generating customers.
- ▪Some researchers are leaving Google to pursue opportunities at startups that provide more freedom and resources for AI development.
Opening excerpt (first ~120 words) tap to expand
A woman walks by a giant screen with a Google logo in Paris. (Thibault Camus / Associated Press) By Julia Love May 18, 2026 6:38 AM PT Share via Close extra sharing options Email Facebook X LinkedIn Threads Reddit WhatsApp Copy Link URL Copied! Print Google’s powerful in-house chips and booming cloud business have made computing power so coveted that even its own AI researchers struggle for access, competing with revenue-generating customers and flagship Gemini projects.Inside the merged Google DeepMind lab, scarce TPUs now shape which questions get asked, who gets promoted and how fast work proceeds, pushing scientists toward short-term wins over riskier, experimental ideas.Top researchers are quitting to build startups such as Elorian and ReflectionAI, saying outside firms offer more…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Los Angeles Times.