NVIDIA’s Vera Rubin NVL72 platform, first announced at CES 2026, will begin commercial deployment in the second half of the year.
Nebius, Supermicro, and CoreWeave will be among the first companies to offer this system, designed to support AI workloads that require long-range token processing and high-throughput inference capabilities.
The Rubin NVL72 platform is built for use cases where memory capacity, compute density, and low-latency communication between components are essential.
Its infrastructure is designed to support advanced AI workloads with high computational and memory demands.
Why It Matters: Vera Rubin NVL72 enables configurations that are constructed to support training and inference for models that use routing logic, sustained memory bandwidth, and extended context processing. This introduction expands available options for organizations that require stable infrastructure to support AI.
- Nebius Integration Into U.S. and European Infrastructure: Nebius will begin offering the NVIDIA Rubin NVL72 system through its AI Cloud platform and the Nebius Token Factory service. The company will provide direct hardware access and a configuration optimized for post-training and inference workloads. By offering the Rubin system alongside previously deployed NVIDIA GB200 and Grace Blackwell systems, Nebius will support a wider range of compute environments. The infrastructure will be delivered through Nebius data centers located in the United States and Europe, providing customers with localized access and network control. Customers using Nebius will be able to operate open-source models and proprietary AI systems that require a predictable environment.
- Supermicro Expands Manufacturing and Capabilities for Rubin NVL72 and HGX Rubin NVL8: Supermicro will produce hardware systems configured for the NVIDIA Rubin NVL72 and HGX Rubin NVL8 using its U.S.-based manufacturing facilities and in-house design framework. The company is focusing on systems with advanced cooling infrastructure and high-density components. The main system from Supermicro will combine dozens of NVIDIA chips into one cabinet, designed to handle very demanding AI tasks. It uses an updated design that keeps the hardware cool using warm-water technology, helping to manage energy use.
- CoreWeave To Deploy Using Proprietary Orchestration and Diagnostics Platform: CoreWeave plans to add NVIDIA Rubin NVL72 racks to its AI infrastructure during the second half of 2026. The company will manage these systems using its Kubernetes-native Rack Lifecycle Controller, which configures and validates each rack as a single unit before assigning workloads. The deployment will also rely on CoreWeave Mission Control, an operating framework that monitors system health and resource availability across each level of the hardware stack. Rubin-enabled infrastructure will be used for projects that involve inference, large context memory processing, and model training. The platform will support customers working on research applications in fields such as robotics, climate simulation, and biomedical modeling.
- Rubin Includes New CPUs, Memory Features, and Secure Computing Environments: The NVIDIA Vera Rubin platform introduces a set of hardware components and improved connections between GPUs and CPUs. It features NVLink 6 communication fabric, LPDDR5X memory with increased capacity, and confidential computing support for data and model isolation. Systems built with the Rubin platform can operate workloads that require extended context length and narrower compute precision. These configurations are designed for inference pipelines that utilize routing mechanisms and model architectures that surpass the capabilities of previous generations. Performance specifications include over 1.4 petabytes per second of memory bandwidth and more than 75 terabytes of memory across a full rack installation.
- Multiple Deployment Approaches: Customers accessing the Rubin NVL72 platform through Nebius, Supermicro, or CoreWeave will be able to select different entry points depending on how they plan to manage their workloads. Nebius offers direct hardware access and token-based inference APIs, while Supermicro delivers physical systems through its reseller and integrator channels. CoreWeave provides a managed environment with integrated security and monitoring services for production AI. These deployment methods are intended to support a wide range of users who require reliable systems for running advanced digital tools and processes.
Nebius to Offer NVIDIA Vera Rubin NVL72 in US and Europe From H2 2026 – Business Wire
CoreWeave Extends Its Cloud Platform with NVIDIA Rubin Platform – Business Wire
Trusted insights for technology leaders
Our readers are CIOs, CTOs, and senior IT executives who rely on The National CIO Review for smart, curated takes on the trends shaping the enterprise, from GenAI to cybersecurity and beyond.
Subscribe to our 4x a week newsletter to keep up with the insights that matter.
