OpenAI poaches DeepMind's top vision trio

The AI talent war just hit Google DeepMind right where it lives.

OpenAI confirmed in an internal memo that it has poached three senior computer vision and machine learning engineers from DeepMind: Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. The trio will join a newly established OpenAI outpost in Zurich, Switzerland, focusing on multimodal AI models that bridge text, images, and audio.

Setting up shop on Google's turf

All three researchers already live in Zurich and have a long history of publishing together. Beyer, who has over 70,000 followers on X, has never been shy about commenting on OpenAI from the sidelines. When Sam Altman was briefly ousted last year, Beyer publicly posted that the most sensible explanation was Altman having his hands in too many other startups.

Now, he is reporting for duty under Altman. And OpenAI did not even ask the team to relocate to San Francisco. Instead, the company simply opened a new office in Zurich.

Zurich has quietly turned into Europe's premier AI battleground, anchored by the prestigious ETH Zurich university. Google has had a massive engineering presence there for years, but rivals are circling. Earlier this year, the Financial Times reported that Apple poached multiple Google AI researchers for a secretive Zurich laboratory. OpenAI is following the exact same playbook, as part of a global footprint expansion that includes upcoming offices in New York, Seattle, Brussels, Paris, and Singapore.

Poach for poach

Top AI researchers routinely command compensation packages nearing or exceeding seven figures, and the revolving door is spinning faster than ever.

This hire looks an awful lot like payback. Just recently, Tim Brooks, the co-lead behind OpenAI's unreleased video generator Sora, departed to join DeepMind. Snagging three veteran vision researchers from DeepMind gives OpenAI an immediate counterpunch.

It also brings much-needed relief to OpenAI's research bench. The company has suffered severe leadership brain drain in recent months. Co-founder and former chief scientist Ilya Sutskever left to launch Safe Superintelligence, while former CTO Mira Murati departed in September to fundraise for her own stealth startup. Other researchers have steadily leaked to Anthropic and DeepMind.

Across the industry, tech giants are doing whatever it takes to hoard talent. Microsoft effectively acquired Inflection AI by hiring Mustafa Suleyman and most of his staff in March. Google reportedly shelled out $2.7 billion just to bring Character.AI founder Noam Shazeer back under its roof.

Why it matters

We think OpenAI's smartest move here was poaching an existing, tightly-knit squad rather than chasing lone wolves.

Top-tier multimodal research relies heavily on team chemistry. Beyer, Kolesnikov, and Zhai already operate like a well-oiled machine. By opening a local hub in Zurich instead of forcing a San Francisco relocation, OpenAI minimized friction and can put this trio to work on multimodal architectures immediately.

For the broader industry, this move underscores just how critical multimodal capabilities have become. OpenAI may have pioneered consumer AI with ChatGPT and DALL-E, but with Sora still behind closed doors and DeepMind pushing multimodal Gemini hard, the pressure is on. Losing Tim Brooks was a blow; bringing in three heavyweights helps steady the ship.

That said, this kind of mercenary talent churn carries serious long-term risks. When compensation easily clears seven figures and researchers bounce between Mountain View and San Francisco every few quarters, retention becomes a never-ending bidding war. If culture and mission cannot keep researchers in place, OpenAI might find that the doors in Zurich swing both ways.