Anyscale

Anyscale is a San Francisco-based company that develops and commercializes Ray, an open-source Python framework for distributed computing that originated at UC Berkeley's RISELab. Founded in 2019 by the RISELab team that built Ray, including Robert Nishihara, Ion Stoica, and Philipp Moritz, the company packages Ray into a managed cloud platform used by engineering teams to train, fine-tune, and serve large-scale AI and machine learning models across GPU clusters on AWS, Google Cloud, Microsoft Azure, Nebius, and CoreWeave. The Ray project itself has surpassed 500 million downloads and more than 41,000 GitHub stars, with over 1,200 external contributors, making it one of the most widely adopted open-source distributed compute engines in the AI industry.

The Anyscale platform provides tooling for distributed model training with elastic autoscaling, multimodal data curation and processing pipelines, batch embedding generation, and post-training workloads such as reinforcement learning and reinforcement learning from human feedback, using frameworks including vLLM, SkyRL, and veRL that run on top of Ray. Customers use the platform to move machine learning workloads that previously ran on a single machine onto multi-node, multi-GPU clusters without rewriting their code, and to orchestrate jobs across more than one cloud provider from a single control plane.

Anyscale raised a Series C round led by Addition and Intel Capital as part of a broader funding history that included early backing from Andreessen Horowitz, which led its $20.6 million Series A. In July 2026, British AI cloud infrastructure provider Nscale agreed to acquire Anyscale in a deal valued at approximately $1.65 billion, with the stated goal of building a full-stack AI cloud platform that spans chips, data centers, and the software layer for training and serving models. Under the agreement, Anyscale's roughly 200 employees are expected to join Nscale while the company continues to operate under its own brand and serve its existing customer base; Anyscale reported that quarterly revenue had grown 70 percent sequentially ahead of the deal.

Anyscale's headquarters are located at 600 Harrison Street in San Francisco, with a secondary engineering office in Bengaluru, India. The company's target customers span foundation model labs, enterprises running large-scale recommendation and ranking systems, and organizations building custom LLM applications that need to scale beyond single-node infrastructure. While Anyscale is a software and cloud-infrastructure company rather than a maker of physical robots, its Ray-based compute stack is increasingly cited as underlying infrastructure for training the large multimodal and reinforcement-learning models used in robotics foundation models and autonomous systems research, which is why the company is tracked as part of the broader physical-AI and robotics compute supply chain.