Andrew Tulloch is reported to have left Meta for Anthropic before completing one year in his second stint at Meta. The move, reported on September 17, 2026, puts a prominent AI engineer at a rival company during an increasingly intense contest for people who can build and operate large-scale AI systems.

The eye-catching compensation figure attached to the move is disputed. The reported Meta package was valued at $1.5 billion over six years, while Meta called that figure “inaccurate and ridiculous.”

What the reported compensation figure means

The $1.5 billion figure describes the reported value of a six-year Meta package. It is not a confirmed salary figure or a reported cash payment made to Tulloch: Meta disputes the estimate itself.

That distinction matters in technology recruiting. A multiyear package can combine different forms of compensation and conditions, so its headline value describes a projected total rather than a simple amount changing hands.

The career behind the move

Tulloch previously spent 11 years at Facebook and was described as a pivotal engineer on PyTorch, the open-source machine-learning framework. He later co-founded Thinking Machines Lab with OpenAI’s former chief technology officer.

Meta reportedly recruited him from Thinking Machines Lab in 2025 as part of a broader push to expand its AI workforce. Meta also reportedly pursued Thinking Machines Lab before seeking Tulloch individually. The sequence underlines why his return to Meta attracted attention: Tulloch had experience moving between a major technology company, foundational machine-learning software and an AI startup.

Thinking Machines Lab was reported to have raised $2 billion in seed funding. That was the company’s reported financing figure, separate from the compensation package associated with Tulloch’s Meta role.

Why the move matters for AI talent

AI companies are competing for engineers and researchers who understand model development, machine-learning infrastructure and the systems that deliver AI services at scale. A hire with experience across Facebook, PyTorch and an AI startup can strengthen a company’s technical bench without bringing an entire organization with them.

Tulloch’s move also shows why a large compensation headline cannot, by itself, explain how long an engineer will stay at a company. The reported package value and the move to Anthropic are separate facts; the reason for the departure has not been publicly explained by a named actor in the reporting.

What Tulloch is expected to do at Anthropic

At Anthropic, Tulloch is expected to work on the efficiency with which customers access models running in data centers. In practical terms, that concerns the systems involved when customers send requests to AI models and receive results.

The reported focus connects his background in machine-learning software with Anthropic’s work to serve model users efficiently.