Analysis
Google is in advanced talks to pay roughly $1.5 billion to license technology from Mechanize, a 35-person AI coding startup, and to hire members of its model-evaluation staff, according to SiliconANGLE and PYMNTS. The proposed structure would give Google a non-exclusive license to Mechanize's technology rather than the company outright, with some staff joining Google directly -- terms four people familiar with the talks say are still moving.
Mechanize raised just $9.1 million in seed funding earlier this year from investors including former GitHub CEO Nat Friedman, Stripe CEO Patrick Collison and podcaster Dwarkesh Patel, and has spent 2026 training AI systems to write and evaluate code. A roughly $1.5 billion deal for a team that size, less than a year removed from that seed round, would be one of the richest per-employee AI acquihire terms disclosed this year.
“The deal follows a pattern Google has used before with Windsurf and Character AI: license the technology, hire the talent, skip the acquisition.”
The deal follows a pattern Google has used before with Windsurf and Character AI: license the technology, hire the talent, skip the acquisition. Gemini has repeatedly lagged Anthropic's Claude Code and OpenAI's Codex on coding benchmarks and developer adoption, and the talks come weeks after Google's own chief scientist Jeff Dean left to start a competing research venture -- a reminder of how much ground Google's coding tools still have to make up. Neither Google nor Mechanize has commented publicly, and nothing is signed -- the risk for Google is paying acquihire-scale money for a team that hasn't yet shipped a product at Gemini's scale, a bet that only pays off if Mechanize's evaluation methods generalize beyond a 35-person lab.
