Nscale is buying Anyscale for $1.65 billion, folding the Ray-based workload orchestration platform into a British neocloud that now spans energy, data centers, GPU capacity, and the software layer that decides how customers actually use those GPUs. The deal, first reported by Bloomberg, hands Nscale a piece of the stack that most compute providers rent out rather than own. It arrives four months after Nscale raised $2 billion in a Series C at a $14.6 billion valuation.
Anyscale was last valued at $1.38 billion in a 2022 Series C round, meaning the $1.65 billion price represents only a modest step up over nearly four years — a reminder that private infrastructure-software marks compressed while GPU-adjacent hardware valuations exploded. The company said revenue rose 70% quarter over quarter in its most recent period, a growth rate that likely made the sale defensible even at a subdued multiple. All roughly 200 employees are joining Nscale.
Anyscale was founded by the team behind Project Ray, the open-source distributed Python framework that has become a default for running large workloads across clusters. The company's original pitch was general-purpose scale-out compute, but after the launch of GPT-3 in 2022 it pivoted to serving and training large language models, data curation, inference, and reinforcement learning. Its commercial platform bundles developer tools, observability, and orchestration on top of Ray.
Key facts
- 01Nscale is paying $1.65B for Anyscale, absorbing all roughly 200 employees under the Anyscale brand.
- 02Nscale raised $2B in a March Series C at a $14.6B valuation, backed by Nvidia, Nokia, Blue Owl, Dell, and Aker.
- 03Anyscale was last valued at $1.38B in its 2022 Series C, a 1.2x mark relative to today's sale price.
- 04Anyscale said revenue rose 70% quarter over quarter in its most recent quarter.
- 05Nscale has locked in compute and data center deals with Microsoft, British Telecom, and Nordcraft.
For Nscale, the logic is vertical integration. The neocloud already operates business lines across power, data center construction, and GPU orchestration, and has been signing compute and data center partnerships with Microsoft, British Telecom, and Nordcraft. Anyscale slots in above all of that as the workload management and scaling tier customers touch directly.
“Together, Anyscale and Nscale can co-design the software layer and infrastructure beneath it, something that neither company could do as effectively by optimizing its layer alone”— Anyscale, company statement
Anyscale will continue to operate under its own brand and serve existing customers, Nscale said — a structure that lets Anyscale keep multi-cloud users who might not want their orchestration layer owned by a single infrastructure vendor. Whether that independence survives the integration in practice is the open question, since the entire strategic point of the deal is co-design between the two layers.
The transaction fits a broader pattern in the AI infrastructure market: neoclouds are no longer content to be GPU-rental businesses. CoreWeave has been pushing into managed inference services. Together AI and Lambda have built out their own model-serving platforms. Owning the software layer that decides how customers consume GPUs means capturing more of each AI dollar spent and, more importantly, making it harder for those customers to leave.
Nscale's investor list — Nvidia, Nokia, Blue Owl, Dell, and Norwegian industrial group Aker — reflects the same thesis from the capital side. Nvidia in particular has been backing neoclouds that build differentiated software on top of its hardware, since a commodity GPU-rental market ultimately squeezes margins for everyone including Nvidia.
The risk in a deal like this is cultural and technical rather than financial. Ray's credibility rests partly on being genuinely neutral open-source infrastructure that runs anywhere. If enterprise customers start to perceive Anyscale as a channel to lock them onto Nscale hardware, the very asset Nscale just paid $1.65 billion for erodes. Anyscale's continued independent branding is a hedge against that perception, but it only works if the substance follows.
The bigger read is that the AI infrastructure business is consolidating faster than the model business. Frontier labs are still expanding, but the picks-and-shovels tier — power, data centers, GPUs, orchestration, workload scaling — is collapsing into a smaller number of vertically integrated players. Nscale is betting that owning four layers of the stack beats renting one, and at a $14.6 billion valuation with $2 billion in fresh cash, it has the balance sheet to keep buying the pieces it's missing.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
