Mirendil signs Google Cloud compute deal worth upwards of $100 million

The multi-year pact gives the AI startup access to Google TPUs, Nvidia GPUs and managed training clusters for self-improving AI research as cloud providers compete for high-growth labs.

19d ago

Fact Check

The TechCrunch exclusive report directly confirms every element of the claim: the $100M+ multi-year Google Cloud deal, access to Google TPUs, Nvidia GPUs, and managed training clusters, and the self-improving AI research purpose. CryptoBriefing and RootData independently corroborate the same figures and details. Supporting company facts (founders, Anthropic/Google DeepMind background, $200M seed at $1B valuation) are consistent across additional independent sources. No conflicting evidence was found.

Summary

Mirendil has entered a multi-year partnership with Google Cloud that its co-founder and CEO, Benham Neyshabur, said is worth upwards of $100 million. The agreement gives the startup access to Google’s TPUs, Nvidia GPUs and managed training clusters to support work on self-improving AI, also known as recursive self-improvement. The arrangement reflects a broader race in artificial intelligence, with cloud providers offering large infrastructure commitments to promising startups while AI companies lock in as much compute capacity as possible as they scale. Neyshabur said Mirendil ultimately wants to build AI capable of taking on the work of an entire frontier AI lab, and argues the technology could automate significant parts of scientific and AI research across areas such as medicine, biology and materials science. Google, for its part, gains a strategic partner developing frontier recursive self-improving AI that it could eventually market to enterprise customers.

Terms & Concepts

TPUs: Google-designed chips for AI workloads
GPUs: Graphics chips widely used for AI training
recursive self-improvement: AI iteratively improving its own capabilities