CoreWeave, Inc. has announced plans to incorporate NVIDIA Rubin technology into its AI cloud platform. This move is expected to broaden the range of solutions available for customers working on agentic AI, reasoning, and large-scale inference tasks. The company aims to be among the first cloud providers to deploy the NVIDIA Rubin platform in the latter half of 2026.
CoreWeave’s cloud infrastructure is designed to handle large-scale AI across multiple generations of technology, allowing clients to match their workloads with appropriate systems as needs change. The integration of NVIDIA Rubin is intended to boost performance, efficiency, and scalability for enterprises, AI research labs, and startups that run production-level AI workloads.
Michael Intrator, Co-founder, Chairman, and Chief Executive Officer of CoreWeave said: “The NVIDIA Rubin platform represents an important advancement as AI evolves toward more sophisticated reasoning and agentic use cases. Enterprises come to CoreWeave for real choice and the ability to run complex workloads reliably at production scale. With CoreWeave Mission Control as our operating standard, we can bring new technologies like Rubin to market quickly and enable our customers to deploy their innovations at scale with confidence.”
Jensen Huang, Founder and Chief Executive Officer of NVIDIA added: “CoreWeave’s speed, scale, and ingenuity make them an essential partner in this new era of computing. With Rubin, we’re pushing the boundaries of AI—from reasoning to agentic AI —and CoreWeave is helping turn that potential into production as one of the first to deploy it later this year. Together, we’re not just deploying infrastructure—we’re building the AI factories of the future.”
The NVIDIA Rubin platform is built for demanding applications such as agentic AI models, drug discovery projects, genomic research efforts, climate simulations, and fusion energy modeling. It supports large-scale mixture-of-experts models that require extensive compute resources over long periods. On CoreWeave’s platform, these workloads will benefit from increased flexibility and consistent performance.
Previously, CoreWeave was noted as the first cloud provider offering general availability for NVIDIA GB200 NVL72 instances and the Grace Blackwell Ultra NVL72 platform. Its proprietary software stack aims to speed up deployment while maintaining high standards for performance and reliability.
NVIDIA Rubin will operate through CoreWeave Mission Control—an operating standard unifying security measures with expert-led operations and system observability—to ensure reliability and transparency across training, inference, and agentic AI workloads. The system includes real-time diagnostics via integration with NVIDIA’s Reliability, Availability, and Serviceability (RAS) Engine.
To address requirements around power delivery and cooling at scale—as well as network integration—CoreWeave has developed a Kubernetes-native orchestrator called Rack Lifecycle Controller. This tool treats entire NVIDIA Vera Rubin NVL72 racks as single programmable units for coordinated provisioning and validation prior to customer deployment.
Dan O’Brien, President and COO at The Futurum Group commented: “Workloads like drug discovery, climate modeling, and advanced robotics demand both cutting-edge compute and the ability to run it reliably at scale. The NVIDIA Rubin platform expands what is possible, and platforms like CoreWeave are what make those capabilities available in practice. That combination is what accelerates real progress.”
Once integrated into its cloud platform later this year (https://www.coreweave.com/), CoreWeave expects customers will be able to focus on building advanced AI systems without needing to manage infrastructure directly.
The expansion follows recent innovations by CoreWeave—including a publicly available managed reinforcement learning capability—and builds on industry recognition reflected in top rankings from MLPerf benchmarks (https://mlcommons.org/en/news/mlperf-training-v4-0/) as well as Platinum status in SemiAnalysis ClusterMAX evaluations (https://www.semianalysis.com/p/semanalysis-clustermax-20-ai-datacenter).
