Google loses AI talent war as Shazeer and Jumper exit DeepMind

Google DeepMind lost two marquee researchers in a single week. Noam Shazeer — co-author of the 2017 'Attention Is All You Need' paper that introduced the transformer (the 'T' in ChatGPT) — announced he is leaving for OpenAI, even though Google had paid more than $2 billion to acqui-hire him and part of his Character.ai team. Two days later, John Jumper, who shared the 2024 Nobel Prize in Chemistry for AlphaFold, said he is leaving DeepMind for Anthropic.

The departures crystallize an intensifying, AGI-driven talent war at the very top of the field. As Axios noted, even as labs race to automate AI research itself, they are placing 'extraordinary value on a tiny number of humans who know how to direct that work,' and key hires tend to spur further hiring. The same week saw Barret Zoph re-depart OpenAI and Nvidia acqui-hire the Essential AI team including Ashish Vaswani — a churn that observers dubbed AI's 'celebrity era' of hiring.

The timing is awkward for Google, which separately delayed its Gemini 3.5 Pro frontier model from June to July to gather more feedback and tweak the model — even as it did ship computer use in Gemini 3.5 Flash. DeepMind leadership, including Demis Hassabis, has publicly insisted Google is still winning on talent, but two simultaneous frontier-level exits to its two fiercest rivals undercut that narrative.

The skeptical read is that individual departures, however symbolic, don't automatically dent a 1,000-researcher org with deep compute and data advantages. The watch item is whether this becomes a trend — and whether the Gemini 3.5 Pro slip signals deeper roadmap friction or just standard polish.