In a major shift in AI research circles, Meta AI has hired three leading scientists—Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—who previously worked at OpenAI and DeepMind. All three are renowned for their contributions to computer vision, particularly the development of the Vision Transformer (ViT) architecture. Their departure signals another wave of high-profile OpenAI defections. The trio had recently opened OpenAI’s Zurich office in late 2024.
With this hiring, Meta gains researchers with expertise in scalable vision models at a time when competition among tech giants in foundational AI is intensifying. Beyer confirmed their move on X, clarifying that he did not receive the speculated $100 million signing bonus. The figure had circulated after OpenAI CEO Sam Altman referenced high compensation offers in the AI job market, though not directly naming Beyer.
Zuckerberg’s Direct Role in AI Recruitment
Meta CEO Mark Zuckerberg personally led the recruitment effort, signaling the importance of the hires for the company’s AI ambitions. Sources say Meta approached the trio shortly after internal dissatisfaction with the company’s latest model performance. Zuckerberg’s hands-on strategy reflects Meta’s urgency to recover ground in the global AI race.
The company has invested billions in infrastructure, but lagged behind in cutting-edge research breakthroughs compared to rivals like OpenAI and Google DeepMind. Securing top researchers from both firms gives Meta a rare edge in expertise. The rumored compensation packages—up to $100 million—were part of Meta’s broader effort to make offers that top rivals couldn’t match. While Beyer denied such figures, Altman’s earlier comments may have been intended to pressure Meta to overspend or stir public speculation. Either way, the hires show Meta’s willingness to pay for elite AI talent.
Vision Transformer Works to Fuel Meta’s Next AI Models
All three hires are closely associated with the Vision Transformer, a model architecture that revolutionized how AI systems handle visual data. Their work at DeepMind and OpenAI helped scale these models for real-world use, and Meta plans to integrate that knowledge into its next-gen AI systems. Meta AI’s recent models have struggled to meet internal benchmarks. The company now expects Beyer, Kolesnikov, and Zhai to help design more efficient and powerful vision architectures.
Their experience in bridging theoretical advances with production-ready applications aligns with Meta’s push to commercialize AI tools faster. This hiring wave suggests Meta is refocusing on core research instead of only optimizing current models. With ViT architects now in-house, the company gains both technical insight and leadership for future visual model development.
Meta Doubles Down on AI Research Depth
These OpenAI defections also reflect a deeper strategy by Meta to build long-term AI depth. The company recently invested $14 billion in Scale AI, a data infrastructure firm that supports model training and labeling. That deal followed a series of AI infrastructure bets aimed at accelerating foundational research. The hiring of the Vision Transformer creators complements these moves. It shifts Meta’s AI narrative from playing catch-up to becoming a destination for top researchers seeking academic freedom and strong internal support.
The Zurich base will allow the team to operate in a familiar environment while building out Meta’s European AI presence. By acquiring leaders with DeepMind and OpenAI experience, Meta gains a competitive edge in research while weakening its biggest rivals. More importantly, this recruitment indicates that Meta wants to build, not borrow. The company isn’t just seeking shortcuts—it’s investing in talent that can define the next era of vision AI from the ground
About The Author
