San Francisco AI startup Anyscale is set to join Nscale in a deal that could reshape the fast-growing market for AI computing infrastructure. The agreement pairs Anyscale’s software for scaling machine-learning workloads with Nscale’s sprawling mix of data centers, GPUs, and power.
Under the definitive agreement, Anyscale will continue operating under its own brand and serving its existing customers, while its roughly 200 employees across the United States, Europe and India join Nscale, according to Nscale’s announcement. Customers will still be able to choose where they run their workloads, with Nscale offering its infrastructure as an additional option.
Nscale declined to disclose financial terms, but Bloomberg reported a price of about $1.65 billion, Reuters reported. The transaction remains subject to closing conditions and regulatory approvals and is expected to close during the second half of 2026.
A Software Layer For Nscale’s GPU Empire
This is less a chatbot acquisition than an infrastructure marriage. Nscale says it supplies the power, data centers and compute, while Anyscale adds the software layer that helps AI teams process data, train and fine-tune models, run inference and deploy AI agents through workloads spread across thousands of GPUs.
Anyscale was built by the creators of Ray, an open-source framework developed at UC Berkeley for distributed computing. Anyscale's history traces its commercial launch to 2019, while its platform now markets Ray-based tools for data processing, model training and inference across cloud environments.
Why The Deal Matters In San Francisco
Anyscale remains one of the Bay Area’s homegrown AI infrastructure companies, with its headquarters in San Francisco and roots in Berkeley’s research community. Its technology sits below the flashy consumer-facing AI layer, handling the unglamorous but essential work of keeping large-scale workloads moving without forcing every engineering team to reinvent distributed systems.
The deal also lands as Nscale accelerates its American expansion. Hoodline previously reported that Nscale planned a Bellevue engineering hub designed for about 100 workers, while the company said in a July financing announcement that it had closed a $900 million revolving credit facility.
Nscale is competing with AI-focused cloud providers, including CoreWeave and Nebius, Reuters noted. Adding Anyscale gives Nscale a stronger pitch to companies that want not just access to expensive GPUs, but also the software needed to turn that hardware into a production AI system.
If the transaction closes as planned, Nscale will emerge with a broader stack running from energy and physical infrastructure to developer tools. For Anyscale, the agreement offers a path to scale its Ray-based platform alongside a much larger supply of compute without abandoning the brand or customer flexibility that helped build its San Francisco presence.
