Nscale Buys Anyscale for $1.65 Billion, Taking Control of the Ray Framework

London-based AI cloud company Nscale announced on July 24, 2026 that it is acquiring Anyscale, the startup behind Ray, a widely used framework for distributed ML computing. According to Bloomberg, the deal is worth about $1.65 billion. Nscale has not officially disclosed the price and only confirmed the acquisition itself.

What are Anyscale and Ray?

Ray is one of the most widely used open-source frameworks for distributed ML computing. It lets teams scale model training, inference, and data processing on clusters of any size, from a single server to hundreds of nodes. Ray was built as a vendor-neutral tool and runs on AWS, Google Cloud, Azure, and on-premise environments. That technological independence made it a de facto standard among AI teams that need flexible orchestration without being tied to a specific provider.

Anyscale is the commercial company built around Ray, founded by the project's creators. It monetizes the open-source project through the managed Anyscale Platform and enterprise subscriptions that include support, security patches, and extended functionality. Until now, Ray existed as a fully independent tool. Its commercial layer is now moving under the wing of an AI provider with its own infrastructure.

Why did Nscale buy Anyscale for $1.65 billion?

Nscale is an unusual AI cloud provider: it owns its data centers and energy infrastructure rather than renting capacity from hyperscalers. At a time when GPU clusters draw hundreds of megawatts, control over power supply reduces operational risk and the cost of compute. Before the deal, Nscale covered the bottom layer of the stack, physical servers and power. Customers assembled the software layer themselves.

Key facts about the deal:

• Announced on July 24, 2026
• Price of about $1.65 billion according to Bloomberg; Nscale has not officially confirmed it
• Buyer: Nscale (London), an AI cloud with its own data centers and power capacity
• Target: Anyscale, the creators of the Ray framework
• Ray is one of the main open-source tools for distributed ML workloads

The Anyscale acquisition closes that gap. Nscale can now offer a single vertically integrated product, from physical servers and power to ML pipeline orchestration.

Nscale is "buying up the stack," adding a software layer for ML

workloads to its physical infrastructure and power capacity, TNW writes.

For enterprise customers, this means a single SLA and fewer integration points. For Nscale, it means better retention and higher margins from a high-margin software layer.

What changes for Ray developers?

Before the deal, Ray was infrastructure-neutral: people used it regardless of cloud provider. Now that its commercial layer is passing to Nscale, the main question for the ecosystem will be how open the project stays. If the company keeps Ray vendor-neutral, the deal could simply speed up its development. If it starts building lock-in to its own infrastructure, developers and competitors will step up support for alternatives. Acquisitions like this have often been followed by forks of open-source projects, as the community hedges its bets in advance.

The $1.65 billion deal reflects a broader trend: the AI infrastructure market is consolidating faster than expected. Providers that cannot match AWS, Google Cloud, and Azure on scale are choosing vertical integration, using full control of the stack to differentiate themselves and hold on to enterprise AI customers.

What this means

The Anyscale acquisition turns Nscale from a GPU landlord into a full-stack AI operator. The company gets not only software but also access to the Ray developer community, one of the most valuable long-term assets in the AI infrastructure segment. Its success will depend largely on whether Nscale keeps Ray an open standard for the whole market or turns it into an exclusive competitive advantage.