AI startup Mirendil has signed a multi-year agreement worth more than $100 million with Google Cloud to secure computing infrastructure for its self-improving AI research.
The partnership reflects growing competition among AI companies to lock in access to high-performance computing, while cloud providers race to attract promising startups developing next-generation AI models.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
Backing The Next Stage Of AI Research
Mirendil plans to use Google’s Tensor Processing Units (TPUs), Nvidia GPUs and managed training infrastructure to develop AI systems capable of improving their own performance over time.
Known as recursive self-improvement, the concept focuses on building AI that can refine its knowledge and capabilities with minimal human intervention. The technology is attracting growing interest across the industry, with several startups and leading AI labs exploring similar approaches.
According to co-founder and Chief Executive Behnam Neyshabur, the long-term goal is to develop AI that can automate scientific research and accelerate discoveries in fields such as medicine, biology and materials science.
Compute Capacity Becomes A Strategic Asset
Training increasingly advanced AI models requires enormous computing resources, making long-term infrastructure agreements a critical competitive advantage.
Mirendil said Google’s combination of TPUs and GPUs allows workloads to be matched with the most suitable hardware, improving efficiency while reducing costs for customers.
For Google Cloud, the agreement strengthens its position in the race to provide infrastructure for frontier AI developers, while giving the company exposure to one of the industry’s emerging approaches to next-generation artificial intelligence.
