Google has completed its talent acquisition of AI startup Mechanize for more than $1.5 billion, absorbing the team into DeepMind to bolster midtraining work that the company hopes will close the gap with rivals. The deal, previously reported by Business Insider, brings over a dozen employees including co-founder Tamay Besiroglu into Google's AI division, while the startup's former chief of staff Guive Assadi remains at Mechanize with the CEO title.
The structure is the story
The transaction follows Google's established pattern of talent deals structured to avoid the antitrust scrutiny that would accompany a formal acquisition. No consideration type, break fee, or premium to Mechanize's last $500 million valuation has been disclosed. The company raised $9.1 million earlier this year at that valuation, making the reported price a multiple of roughly 165 times the last private round, a figure that reflects the scarcity premium for teams that can move the needle on model performance rather than any revenue multiple.
Midtraining is the bottleneck
Mechanize's specialty is improving AI models at coding, a persistent weakness in Google's lineup. The incoming team is being deployed on midtraining, the phase between pre-training and fine-tuning where models learn to reason and use tools. That is the same layer where Google's Windsurf talent deal last year placed Varun Mohan in charge of the Antigravity agentic coding program. Two deals, same architectural bet: the path to a competitive model runs through better post-training infrastructure.
The counter-narrative
Google and Besiroglu declined to comment on the deal. Assadi's continued presence as CEO of a company whose core team and technology have effectively migrated to DeepMind raises the question of what, if any, independent operations remain. Talent deals of this scale typically leave a shell entity to manage legacy obligations, but the source does not clarify whether Mechanize retains any IP, customers, or revenue stream.
What to watch
The next signal is whether DeepMind's next model release shows measurable gains on coding benchmarks. Google has used talent deals to plug capability gaps before; the test is whether the Mechanize integration produces a step change or merely keeps pace. With no disclosed terms, the market cannot assess the return threshold Google has set for this deployment of capital.