Mirendil secures over $100 million Google Cloud contract for self-improving AI

Mirendil, an AI laboratory focused on recursive self-improvement, has signed a multi-year contract with Google Cloud valued at over $100 million to secure the massive compute resources needed for its research. The agreement provides the startup with both Google’s tensor processing units and Nvidia graphics processors, as well as managed training clusters that can be tailored to specific workloads. Benham Neyshabur, the company’s co-founder and chief executive, noted that the deal equals about half of the capital raised in Mirendil’s recent seed round, which valued the firm at $1 billion.

Harsh Mehta, another co-founder, explained that the ability to allocate the right tasks to the right accelerators reduces costs for both Mirendil and its future customers. Google’s AI infrastructure senior vice president Amin Vahdat said the focus is now on coordinating entire intelligence systems to overcome scaling limits. Mirendil hopes its self-improving AI will eventually conduct autonomous research in areas like Alzheimer’s disease, accelerating progress in medicine, biology and materials science. In return, Google gains a strategic partner developing frontier AI technology it can later offer to enterprise clients.