Anyscale Transfers Ray To PyTorch Foundation

Published: Friday, October 24, 2025 at 6:07 am

PyTorch Foundation Welcomes Ray, Consolidating AI Infrastructure

The PyTorch Foundation has expanded its portfolio by adding Ray, a distributed computing framework, as a hosted project. This move aims to create a unified platform for artificial intelligence (AI) workloads, encompassing data processing, model training, and deployment. The integration of Ray with PyTorch and vLLM under the Linux Foundation's open governance model addresses enterprise concerns about vendor lock-in and fragmented infrastructure.

Ray, originally developed by Anyscale and originating from UC Berkeley's RISELab, has gained significant traction, with over 237 million downloads and adoption by major companies. The PyTorch Foundation's governance structure, which separates technical and business decision-making, ensures that no single entity controls the direction of the projects. A Technical Advisory Council manages technical aspects, while a Governing Board, composed of representatives from leading tech companies, handles business decisions.

This consolidation offers several benefits. It reduces integration overhead by providing a cohesive stack for AI tasks. Ray handles the computational demands that PyTorch and vLLM cannot alone, such as scaling data processing and distributing training workloads. The unified governance and aligned roadmaps of the three projects facilitate interoperability.

The neutral home provided by the PyTorch Foundation supports long-term planning for enterprises. The Linux Foundation's track record in sustaining major infrastructure projects provides stability and transparency. This allows organizations to assess project health independently and participate in governance.

The move also accelerates development by fostering collaboration among contributors from various organizations. This reduces redundant development efforts and allows for direct collaboration on shared infrastructure challenges.

Anyscale, the original developer of Ray, will continue to operate as a commercial entity, offering managed Ray services with enhanced performance and enterprise features. This separation between open-source development and commercial offerings provides enterprises with options, allowing them to deploy Ray directly or purchase managed services. The consolidation under the PyTorch Foundation positions the combined stack as a vendor-neutral alternative to proprietary solutions.

BNN's Perspective: This is a positive development for the AI community. The consolidation of these key technologies under a neutral foundation promotes innovation and reduces the risk of vendor lock-in, which is crucial for the long-term health and growth of the AI ecosystem. It allows for greater collaboration and transparency, ultimately benefiting both developers and end-users.

Keywords: PyTorch Foundation, Ray, distributed computing, AI, open source, vLLM, data processing, model training, inference, vendor lock-in, Linux Foundation, Anyscale, governance, infrastructure, ecosystem, collaboration, enterprise, AI workloads, PyTorch, machine learning

Full Story