The battle beyond models: How AI’s top labs are fighting for talent

Google spent roughly $2.7 billion in 2024 to bring AI pioneer Noam Shazeer back into its orbit through a deal with Character.AI, the startup he co-founded after leaving the company. Less than two years later, Shazeer is headed to OpenAI.

A day after news of his departure surfaced, another major name followed. John Jumper, the Nobel Prize-winning scientist behind AlphaFold, announced he was leaving Google DeepMind for Anthropic.

This year, much like football players changing teams mid-season, AI labs are poaching top talent to develop the most advanced and sophisticated AI systems.

Google's high-profile departures

Shazeer is among the most influential figures in modern AI. He was one of the co-authors of the landmark 2017 paper Attention Is All You Need, which introduced the transformer architecture that underpins systems such as ChatGPT, Gemini and Claude.

His ties with Google stretch back decades. After leaving the company to co-found Character.AI, he returned through Google's multi-billion-dollar deal with the startup. Now he is moving again, this time to OpenAI, where he is expected to play a key role in future model development.

Jumper's move is significant for a different reason. While Shazeer helped lay the foundations for today's AI models, Jumper became known for applying AI to science.

He spent nearly nine years at DeepMind and, in 2024, shared the Nobel Prize in Chemistry with DeepMind chief executive Demis Hassabis for AlphaFold. The system transformed protein structure prediction and is widely regarded as one of the most important real-world applications of artificial intelligence so far.

Anthropic's growing appeal

Anthropic has been one of the clearest beneficiaries of this talent movement. In May, the company added another high-profile recruit when Andrej Karpathy joined its pre-training team. Karpathy was part of OpenAI's founding group before going on to lead Tesla's Autopilot programme.

Over the years, he has also become one of the most recognisable educators in AI through his technical explainers, coding projects and online courses. At Anthropic, he is working on the systems that help train future generations of Claude.

In April, Eric Boyd joined as head of infrastructure after spending nearly 17 years at Microsoft, where he most recently served as president of Azure AI. The hire highlights a growing reality in AI: success is no longer determined solely by research breakthroughs. Running and scaling these models has become just as important.

Ross Nordeen's arrival added another notable name to the roster. Nordeen, one of xAI's original co-founders, left the Elon Musk-led company earlier this year before joining Anthropic. His departure marked the end of xAI's founding team, with Musk remaining the only original founder still at the company.

The trend is reflected in hiring data as well. According to SignalFire's State of Talent Report, Anthropic recorded the highest employee retention rate among major frontier AI labs at 80%, ahead of Google DeepMind at 78%, OpenAI at 67%, and Meta at 64%.

The report also found that researchers were far more likely to move from rival labs to Anthropic than the other way around, suggesting the company has become one of the industry's most attractive destinations for AI talent.

OpenAI continues to build its bench

OpenAI has also been active in the talent market. Alongside Shazeer, the company hired Ruoming Pang earlier this year. Pang previously led Apple's foundation models team before being recruited by Meta during Mark Zuckerberg's push to expand the company's AI capabilities.