Whats new in Ray and Anyscale: KubeRay Connect, Native RDMA, and more | Ray Summit 2026 @anyscale
Whats new in Ray and Anyscale: KubeRay Connect, Native RDMA, and more | Ray Summit 2026  @anyscale
Uploaded September 2026 | Updated November 2026, 5 days ago
Anyscale and Nscale are coming together to build a vertically integrated, AI-native computing stack, optimized across every layer from power to tokens. That news opens the Day 1 keynote, followed by every Ray and Anyscale announcement of the summit.

Anyscale CEO Keerti Melkote makes the case for owning your intelligence through continual learning loops, then sits down with Josh Payne, founder and CEO of Nscale, to talk through what the combination means for Ray, for Anyscale's multi-cloud strategy, and for the developers building on both, including Nscale's commitment of over 300,000 GPUs over the next two years and Ray remaining open source and community-governed under the PyTorch Foundation.

Robert Nishihara, creator of Ray, follows with the anatomy of a reinforcement learning workload and why it is one of the hardest distributed systems problems in AI. The Ray team then covers a year of development: cgroup-based resource isolation on Kubernetes, the new Ray history server, 10x faster startup for 32,000 actors, topology-aware scheduling for multi-host TPUs and NVIDIA Blackwell GPUs, Ray Direct Transport in beta, Ray Data scaling to 5,000 GPUs, Ray Sandboxes launching 100,000 sandboxes in 30 seconds, and Ray Serve LLM reaching general availability. Christian Stano, Field CTO, maps the platform stack from one cluster to many clouds, followed by a live demo of coding agents building a complete learning loop and two announcements: Anyscale KubeRay Connect, which connects an existing KubeRay cluster to Anyscale in minutes with no migration, and Anyscale GPU health observability, which takes hardware root cause analysis from hours to seconds.

Chapters:
00:00 Intro: Keerti Melkote, CEO, Anyscale
07:28 Nscale Vision and Partnership: Josh Payne, founder and CEO, Nscale
14:34 Anatomy of AI Workloads: Robert Nishihara, creator of Ray and co-founder, Anyscale
24:39 Ray Updates: Reliability and Scale: Edward Oakes, Staff Software Engineer, Anyscale
33:26 Ray Libraries and Workloads: Richard Liaw, Product Manager, Anyscale
40:56 Platform Challenges and Scaling: Christian Stano, Field CTO, Anyscale, Goku Mohandas, Product Manager, and Kourosh Hakhamaneshi, Team Lead, Anyscale
55:10 Kubernetes and Observability: Elizabeth Hu, Product Manager, Anyscale and Mike Tower, Product Manager, Anyscale

Liked this video? Check out other Ray Summit keynote sessions: youtube.com/playlist?list=PLYnBpswCtPo4

Subscribe to our YouTube channel to stay up-to-date on the future of AI: youtube.com/c/anyscale

đź”— Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute
Website: anyscale.com
Whats new in Ray and Anyscale: KubeRay Connect, Native RDMA, and more | Ray Summit 2026Anyscale on Azure: Build and deploy AI at scale in your own tenantWhy Current AI Cannot Truly See the World | Elorian AI | Ray Summit 2026Prompt Learning: A Reinforcement Learning-Inspired Approach to AI Optimization | Ray Summit 2025Introducing Terminal-Bench: Evaluating LLM Agents in Realistic Terminal Settings | Ray Summit 2025
Anyscale |

What's new in Ray and Anyscale: KubeRay Connect, Native RDMA, and more | Ray Summit 2026

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER