Nvidia-Backed Nscale to Acquire Anyscale, the Company Behind Ray

British AI cloud company Nscale announced on July 30, 2026 that it has signed a definitive agreement to acquire Anyscale¹. Anyscale provides a platform for scaling AI workloads — data processing, training, inference, and reinforcement learning — across thousands of GPUs, and was founded by the team that created the open-source distributed framework Ray¹².

Financial terms were not officially disclosed¹, but according to TechCrunch, Bloomberg reported the price as $1.65 billion, citing an anonymous source². The deal is expected to close in the second half of 2026, subject to closing conditions and regulatory approvals¹.

A GPU-and-Power Company Buys the Software Layer

Nscale is a neocloud that pitches a “full-stack AI cloud” vertically integrating power, data centers, compute, and software; its investors include NVIDIA, Nokia, Blue Owl, Dell, and Norway’s Aker¹². The company raised a $2 billion Series C in March 2026 at a $14.6 billion valuation, and its customers and partners include Microsoft and British Telecom².

Anyscale, for its part, was founded to commercialize Ray, which began as an open-source project out of UC Berkeley. Ray was donated to the PyTorch Foundation in 2025 and remains open source under community governance¹. Anyscale itself is growing fast, with revenue up 70% quarter-over-quarter in its latest quarter². Its team of roughly 200 people across the United States, Europe, and India will join Nscale, while the Anyscale brand and service to existing customers continue¹.

Nscale CEO Josh Payne framed the deal as completing the company’s offering at the software layer, and Anyscale CEO Keerti Melkote described the goal as creating “the first full-stack AI hyperscaler”¹. TechCrunch relays the companies’ explanation that together they can “co-design the software layer and infrastructure beneath it”².

After the Race to Secure Compute Comes the Race to Stack It

This acquisition extends a run of July moves around compute infrastructure. Sutskever’s SSI announced a long-term strategic partnership with NVIDIA to scale its compute by an order of magnitude, and Recursive signed a $410 million compute deal with AWS. New financing structures are appearing too, like loans collateralized by inference chips. While research labs compete to secure compute, the suppliers are finding that GPUs and power alone no longer differentiate — so they are moving up the stack to own the software that runs on top.

A company commercializing a well-known open-source framework for distributed machine learning becoming part of a GPU cloud could shift the boundary between “neutral software” and “a specific cloud’s asset” in AI infrastructure. That said, Ray itself remains open source under the PyTorch Foundation¹, so developers using Ray are not immediately affected. What deserves watching is whether Anyscale’s commercial platform becomes optimized for Nscale’s infrastructure over time, and how much freedom customers retain in choosing where their workloads run.

If you are deciding which cloud to build your AI platform on, note that neoclouds are shifting their pitch from “cheap GPUs” to “a full stack including operations software.” Vendor evaluation now needs to weigh not just price but potential lock-in at the workload-management layer.