Nscale has acquired Anyscale for approximately $1.65 billion to advance its position in the AI sector.

      Nscale, the London-based AI-cloud firm that operates its own power and data centers, is expanding its operations by acquiring Anyscale, the startup responsible for the popular Ray software. According to Bloomberg, the acquisition is valued at approximately $1.65 billion.

      Nscale announced the acquisition on Thursday but did not disclose a purchase price; Bloomberg reported the figure based on information from a source familiar with the transaction. Anyscale, which has around 200 employees located in the US, Europe, and India, will become part of Nscale, with the deal expected to finalize in the latter half of 2026.

      Nscale's strategy stems from its ownership of the foundational layers beneath the chip—electricity, infrastructure, and hardware. By acquiring Anyscale, it is advancing into software that optimizes the work efficiency of each costly GPU.

      The rationale centers on profit margins. Renting raw GPUs hourly is a commodity business, with fierce competition among companies acquiring the same Nvidia systems. More value is generated in how tasks are executed across numerous chips simultaneously.

      Anyscale's platform, built on the open-source Ray framework, effectively manages data preparation, training, fine-tuning, and inference across a fleet of resources. The company claims that its solution can reduce the total cost of ownership by as much as 90% compared to using disparate tools.

      “Most infrastructure providers purely buy GPUs and lease them,” stated Josh Payne, Nscale’s CEO and founder. He emphasized that Nscale differentiates itself by building and owning every component: the energy sources, data centers, computing resources, and software.

      Dan Bathurst, the chief product officer, clarified further that engineers can now consider Nscale as “a one-stop shop for all their training, fine-tuning, and inference services.”

      Nscale is positioned among the “neoclouds,” a new wave of data-center companies focused on AI that emerged from a cryptocurrency-mining operation in early 2024. It has rapidly expanded, surpassing 1GW of capacity and securing significant computing contracts.

      The company is also in a strong financial position for continued growth, having secured a $900 million revolving credit facility in July, along with a $2.5 billion commitment for UK data centers. Plans to go public may occur later this year, and its board includes former Meta executives Sheryl Sandberg and Nick Clegg.

      Competitors are making similar moves. In May, Nebius invested $643 million to acquire Eigen AI, a firm that helps optimize AI operations. Additionally, the chip manufacturer Qualcomm recently purchased the compiler startup Modular. “Everyone is looking to move up the stack,” Bathurst told Bloomberg, highlighting the goal of becoming a comprehensive provider as investment in AI infrastructure accelerates.

      What Nscale is truly acquiring is the commercial platform, the engineering talent, and the client base—comprising companies such as Coinbase, Runway, and Bedrock Robotics—while also joining the PyTorch Foundation. Anyscale reportedly experienced a 70% revenue growth in its latest quarter, with CEO Keerti Melkote dubbing the new entity “the first full-stack AI hyperscaler.” This strategy aims to consolidate control over all aspects, from energy supply to the trained models, paralleling the drive to establish national AI factories.

      The success of this endeavor hinges on timing. Anyscale currently operates based on available capacity, but Nscale’s largest deployment, a 1.35GW facility for Microsoft in West Virginia, will not begin operations until 2027.