Google DeepMind Haemorrhages Talent as AI Startups Outbid Big Tech for Top Researchers
Google DeepMind Haemorrhages Talent as AI Startups Outbid Big Tech for Top Researchers
Alphabet’s Google DeepMind is facing a deepening talent crisis, losing several of its most prominent researchers to rivals OpenAI and Anthropic in rapid succession — exits that wiped roughly $270 billion from Alphabet’s market capitalisation on the NASDAQ in a single selloff. The departures signal a structural shift in where elite AI researchers believe the most consequential work is now being done.
The Google Exodus
The most high-profile departure came on June 20, when Noam Shazeer — Google’s vice president of engineering and co-lead of its Gemini models — announced he was joining OpenAI. The move is particularly costly for Google: the company had spent approximately $2.7 billion less than two years ago to reacquire Shazeer and his team after he had left to co-found Character.AI.
Shazeer is no ordinary engineer. He joined Google in 2000 and co-authored the landmark 2017 paper Attention Is All You Need, whose transformer architecture underpins virtually every major large language model in use today, including Gemini itself.
His exit followed by just three days the departure of Nobel Laureate John Jumper, who left Google’s AI lab for Anthropic. Bloomberg subsequently reported that researchers Jonas Adler and Alexander Pritzel are also expected to follow Jumper to Anthropic.
OpenAI Builds Out Research, Policy, and Agent Capabilities
OpenAI has been the primary beneficiary of the talent reshuffling. Beyond Shazeer, the Sam Altman-led organisation hired Dean Ball, who helped shape AI policy in the early months of the Trump administration. Ball will lead OpenAI’s newly formed Strategic Futures team, focused on frontier AI policy.
In February, OpenAI brought on Peter Steinberger, creator of the open-source AI agent framework OpenClaw, to develop its next generation of autonomous agents. Steinberger publicly disclosed that he turned down a higher financial offer from Meta, citing stronger alignment with OpenAI’s product philosophy and its stated goal of building broadly accessible AI systems.
According to AI talent tracker Trueup.io, OpenAI has recorded 17 senior arrivals against 12 departures since the start of 2025 — a net positive that few incumbents can match.
Anthropic’s Gravitational Pull Grows Stronger
Anthropic has emerged as perhaps the most aggressive acquirer of senior talent. In May, the company recruited Andrej Karpathy — OpenAI co-founder and former head of AI at Tesla — to its pre-training division, where he is building a team focused on using Claude to accelerate future model research.
The startup also hired Eric Boyd, a 16-year Microsoft veteran who led the company’s AI platform and oversaw large-scale model deployments. Boyd now heads Anthropic’s infrastructure division.
The numbers reinforce Anthropic’s growing pull. A 2025 analysis by venture capital firm SignalFire found that Google DeepMind engineers were approximately 11 times more likely to move to Anthropic than the reverse. Trueup.io data shows Anthropic recording 9 senior executive arrivals against 4 departures so far this year; since 2024, the ratio stands at 18 arrivals to 9 departures.
In an interview with Bloomberg, Anthropic co-founder and CEO Dario Amodei pointed to the startup’s founding team stability as a differentiator, noting that Anthropic remains the only leading AI lab to have retained all of its co-founders — a claim that carries weight in an industry where defections have become routine.
What the Talent Flow Reveals
The pattern emerging from these moves is not simply about compensation. Steinberger’s decision to accept less money from OpenAI over a richer offer from Meta, and Karpathy’s pivot from Tesla to a safety-focused startup, suggest that researchers are increasingly weighing mission alignment and technical autonomy alongside financial incentives.
For Google, the strategic cost is compounding. Having spent billions repatriating talent in recent years, it now faces the prospect that neither money nor institutional scale is sufficient to anchor researchers who believe the frontier has shifted elsewhere.
