Nscale acquires Anyscale for $1.65B to control full AI compute stack
In brief
- Nscale acquired Anyscale for $1.65 billion, announced July 30 and expected to close in H2 2026
- Anyscale commercializes Ray, an open-source distributed computing framework originally developed at UC Berkeley
- Deal combines Nscale's data centers and power infrastructure with Ray's workload orchestration software
- Ray enables companies to reduce deployment costs by up to 90% in some cases
- Combined entity controls full stack from electricity powering GPUs to software assigning tasks
The stack gets tighter
Anyscale is the commercial company behind Ray, which handles the orchestration of AI workloads across massive GPU clusters. Ray spreads data processing, model training, inference, and reinforcement learning tasks in ways that companies can reportedly cut deployment costs by as much as 90% in some cases.
The acquisition marries two layers of the AI stack. Nscale already owns its own data centers and power generation infrastructure. Adding Ray means the combined entity controls everything from the electricity powering the GPUs to the software deciding which GPU handles which task.
What Nscale gets
Anyscale brings roughly 200 employees to the combined entity. The deal gives Nscale a proven, open-source framework with deep roots in the AI research community—Ray was originally developed at UC Berkeley—plus a team that's spent years optimizing how distributed workloads run at scale.
Vertical integration in infrastructure has become a competitive advantage in AI. By owning both the hardware layer and the orchestration layer, Nscale can optimize for efficiency in ways that pure software or pure infrastructure players can't. The deal positions the company to compete on cost, reliability, and performance all at once.
