Nvidia announced that its next-generation Vera Rubin AI platform has entered full production, marking a major moment for the company at the Consumer Electronics Show (CES) 2026 in Las Vegas. The announcement places Nvidia at the center of growing competition in artificial intelligence hardware as companies race to build faster, more efficient systems for training and running large AI models.
Speaking during a CES presentation, Nvidia CEO Jensen Huang said the Vera Rubin chips are now being produced and are expected to begin shipping to customers later this year. Wider deployment of systems based on the platform is planned for the second half of 2026. Nvidia says the new platform is designed to significantly improve performance while lowering the cost of AI computing in data centers.
A New Platform Built for Large-Scale AI
Vera Rubin is not a single chip but a complete AI computing platform made up of six tightly integrated components. At its core are the Rubin GPU and the Vera CPU, which are designed to work together with high-speed networking and data movement technologies. Nvidia says this design allows AI workloads to run more efficiently than on previous systems.
According to the company, the platform can reduce the cost of AI inference to roughly one-tenth of what is required on Nvidia’s current Blackwell-based systems. Training large AI models may also require far fewer chips, potentially cutting hardware needs to about one-quarter compared with earlier architectures.
The platform also includes Nvidia’s latest interconnect and networking technologies, such as updated NVLink switching, data processing units, and Ethernet switches. These components are intended to move massive amounts of data quickly between chips and servers, which is critical for advanced AI workloads.
Rack-Scale Systems and AI Supercomputers
One of the key systems built on the platform is the Vera Rubin NVL72 rack. Nvidia says a single NVL72 rack can contain up to 72 GPUs and 36 CPUs working together as a unified system. Multiple racks can be combined to form a DGX SuperPOD, creating what Nvidia describes as an AI supercomputer for large data centers.
The company says these rack-scale systems can train advanced AI models using far fewer GPUs than previous generations. Nvidia also claims the platform can lower the cost per token—a common measure of efficiency for large language models—by up to ten times compared with earlier systems.
Production Milestone and Timeline
Nvidia confirmed that full production of the Vera Rubin chips is now underway. While the company did not define production volumes, the announcement signals that the platform has moved beyond development and into manufacturing at scale.
Nvidia expects initial systems to reach customers later in 2026, with broader availability in the second half of the year. The company has said that major cloud service providers and AI infrastructure companies are expected to be among the early adopters, deploying Rubin-based systems in large data centers.
Competition Heats Up in AI Hardware
The Vera Rubin announcement comes as competition in the AI chip market continues to intensify. Rivals are developing their own processors, and some large technology companies are investing in custom-designed AI chips to reduce reliance on outside suppliers.
By moving the Rubin platform into production and highlighting its efficiency gains, Nvidia is signaling its intent to maintain leadership in high-performance AI computing. The company is positioning Vera Rubin as a foundation for the next wave of AI development, from large language models to complex data analysis tasks.
As demand for AI computing power continues to grow, Nvidia’s latest platform reflects the company’s strategy to deliver larger, faster, and more cost-effective systems for the world’s biggest AI workloads.
