In an exclusive development reported by TechCrunch, AI laboratory Mirendil has forged a multi-year alliance with Google Cloud to secure the crucial computing power necessary for its pioneering research into self-improving artificial intelligence.
This agreement underscores two significant trends currently shaping the AI sector: major cloud providers are actively engaging startups with substantial infrastructure commitments, while AI companies are aggressively pursuing compute deals to guarantee access as they scale their operations.
Benham Neyshabur, Mirendil’s co-founder and CEO, informed TechCrunch that the partnership is valued at over $100 million. This figure represents approximately half of the seed funding Mirendil successfully raised in late June, which valued the company at $1 billion.
The deal grants the startup access to both Google’s custom Tensor Processing Units (TPUs) and Nvidia’s Graphics Processing Units (GPUs), alongside managed training clusters. These resources will empower Mirendil to advance its self-improving AI, with the ambitious goal that its AI will eventually be capable of performing the work of an entire frontier AI laboratory.
Self-improving AI, also known as recursive self-improvement, describes AI systems designed to enhance their own capabilities iteratively. This advanced concept is a focus for leading labs such as Anthropic, from which Mirendil’s co-founders originate. A number of emerging startups, including Recursive Superintelligence and Ricursive Intelligence, have also recently dedicated their efforts to achieving this objective.
Mirendil envisions this iterative improvement process as a catalyst for automating a significant portion of scientific and AI research, thereby accelerating progress for scientists in critical fields like medicine, biology, and materials science.
Neyshabur articulated his belief that AI can emulate how human scientists acquire knowledge in new domains, accumulate expertise, and progressively refine their performance. He stated, “You can have a self-improving AI where you can point a problem at it and it keeps getting better with time.”
He further elaborated, “How can we have an AI system that keeps doing research, keeps improving its own knowledge and performance when it comes to Alzheimer’s disease?” Neyshabur concluded, “This technology allows us to set goals that are ambitious for AI, and the AI would keep making progress.”
However, training self-improving AI demands immense computational resources. Harsh Mehta, another co-founder of Mirendil, noted that effective training increasingly revolves around precisely matching the appropriate workloads with the optimal hardware.
Mehta explained, “These models are really good at working with different workloads and chips, and assigning the right workloads to the right chips.” He added that Google “provides multiple kinds of chips […] this flexibility allows us to ultimately mix and match workloads with the right kind of accelerators, and then lower the cost not just for us, but also for our customers using our systems.”
This emphasis on flexibility is a cornerstone of Google’s AI infrastructure strategy. Amin Vahdat, Google’s SVP and chief technologist of AI and infrastructure, commented in a statement that AI advancement transcends mere chip-level performance. Instead, it is about “how we orchestrate entire systems of intelligence and break through the physical constraints of scaling.”
Neyshabur highlighted that Mirendil’s proprietary software and systems layer empower customers to maximize the utility of Google’s hardware, potentially providing the cloud giant with a competitive edge. In return, Google gains a strategic partner at the forefront of building recursive self-improving AI—a technology it can eventually offer to its enterprise clients.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.