Co-developed by Anyscale and Microsoft, a new fully managed Microsoft Azure service—powered by Ray—delivers high performance, flexibility, and cloud-scale AI computing for enterprises.
San Francisco, CA – November 4, 2025 — Anyscale today announced a partnership with Microsoft and the private preview of a new AI-native compute service, co-developed with Microsoft and delivered as a fully managed, first-party offering on Microsoft Azure. Powered by Ray, the open source distributed compute framework for AI, this new Azure service provides enterprises with a streamlined, high-performance experience for building, running, and scaling advanced AI workloads.
As organizations increasingly embed AI into every customer interaction, they face the limitations of legacy computing systems not designed for today’s dynamic, large scale AI workloads. Modern AI requires high-performance computing that can scale dynamically across diverse data, models, and accelerators—capabilities that traditional computing architectures struggle to deliver efficiently.
The new Anyscale-powered service on Azure addresses these challenges by bringing AI-native computing to every enterprise. Built on open-source Ray and optimized by Anyscale, the jointly engineered service combines the performance and flexibility of Ray’s distributed compute engine, Anyscale’s developer tooling and cluster management expertise, and the security, scale, and trust of Azure.
“AI is redefining every industry, but scaling it remains one of the hardest challenges,” said Keerti Melkote, CEO of Anyscale. “Together with Microsoft, we’re making it dramatically easier for enterprises to build and run AI at scale — combining Anyscale’s platform purpose-built for Ray with Azure’s trusted infrastructure so enterprises can focus on innovation and not on infrastructure management”
Ray: The Foundation for AI-Native Computing
AI workloads have outgrown traditional infrastructure. What starts as a simple experiment on one machine quickly becomes a web of multimodal data pipelines, model training jobs, and inference services that must scale across thousands of GPUs and CPUs – resulting in complexity that slows innovation and AI in production.
Ray, created by the team behind Anyscale, is the open-source engine for AI-native computing—designed to run every stage of the AI lifecycle, from data processing to training and serving, using any Python framework, with any data modality, on any hardware. Ray replaces layers of custom orchestration with a unified framework that efficiently scales processing from one to thousands of machines, enabling faster development and higher resource utilization.
This matters because AI isn’t just bigger, it’s different. Modern workloads mix CPUs and specialized accelerators such as GPUs in the same pipeline – think document processing, generative agents or multimodal models that need to process data to see, hear and speak. Traditional systems struggle to coordinate these resources efficiently, especially with more complex, unstructured data types. Ray was built from the ground up to do exactly that which turns into faster development and higher GPU utilization.
With more than 27 million monthly downloads and 39,000 GitHub stars, Ray has become a foundational technology for AI infrastructure at leading companies like Uber, Spotify, Canva, and Coinbase. With this new Azure service, it is moving from open-source success to now available at enterprise scale.
“Enterprises want flexibility and control as they scale AI workloads. The open source Ray project provides a great choice for distributed AI computing. By bringing Ray to Azure Kubernetes Service through Anyscale on Azure, we’re giving developers the ability to use familiar cloud-native patterns to build and scale AI applications,” said Brendan Burns, Corporate Vice President, Cloud-Native Compute at Microsoft and co-creator of Kubernetes.
Anyscale Running on Azure: AI-Native Compute Democratized
Anyscale running on Azure turns Ray’s powerful, AI-native computing framework into a fully-managed, first-party service. It streamlines setup and development, removes the operational complexity of managing clusters, and accelerates workloads with the Anyscale Runtime – a Ray-compatible, performance-optimized engine – all within Azure’s secure and trusted infrastructure.
As a first-party Microsoft product, the Anyscale managed service on Azure provides AI teams with:
Developer Velocity: Development teams can quickly deploy clusters from Azure Portal, run interactive development with cluster-backed IDE, and debug distributed applications with advanced dashboards.
Production Resilience: Fully managed, fault-tolerant Ray clusters for both batch and low-latency serving directly inside Azure customers’ Azure Kubernetes Service (AKS).
Cost-Efficient Processing: Anyscale Runtime delivers up to 10x faster performance compared to self-managed Ray OSS without code changes
Security and Governance: Anyscale runs directly inside a customer’s Azure account, ensuring data, compute, and AI models remain under their full control and compliance with Azure security policies.
Azure customers can set up and manage Anyscale directly from the Azure Portal, run their Ray-powered AI workloads on Azure Kubernetes Service (AKS), and take advantage of unified billing through Azure. The collaboration enables every Azure developer to build and run – without operational complexity – any AI workload or form of AI from traditional ML to agentic AI.
A New Era of Computing
The partnership between Microsoft and Anyscale marks a new era of computing purpose-built for AI. By combining the power of Ray available as a platform built by Anyscale, the flexibility of Kubernetes, and the trust of Azure infrastructure, this new offering empowers engineering teams to operationalize AI at enterprise scale, accelerating innovation from prototype to production without the complexity of managing distributed systems.
Availability
The new Anyscale service on Azure will enter private preview on November 4, 2025, and will be accessible directly through the Azure Portal. General availability is expected in 2026. Learn more here.
About Anyscale
Anyscale, founded by the creators of Ray, is pioneering the era of AI-native computing. Its platform enables developers and enterprises to easily build, run, and scale AI workloads—from multimodal data processing to training and inference—optimized for modern accelerators. With Anyscale, AI teams get the fastest, most reliable Ray experience to power the next generation of AI applications and platforms. Learn more at www.anyscale.com.