AI lab Mirendil has signed a multi-year deal worth over $100 million with Google Cloud to secure computing capacity for its research into self-improving AI, TechCrunch has learned exclusively. The agreement reflects two broader industry trends: cloud providers offering major infrastructure commitments to attract AI startups, and AI firms racing to lock in compute access as they scale their operations.

The deal, confirmed by Mirendil co-founder and CEO Benham Neyshabur, is worth roughly half of what the startup raised in its seed round at a $1 billion valuation in late June. It gives Mirendil access to Google's TPUs and Nvidia GPUs alongside managed training clusters, which the company will use to develop AI capable of "recursive self-improvement" — systems that iteratively enhance their own performance. Mirendil, founded by former Anthropic researchers, believes the technology could automate scientific research in fields such as medicine, biology and materials science, while Google gains a strategic partner developing frontier AI it could eventually offer to enterprise customers.