Nvidia unveiled its Vera Rubin computing platform Monday at the company's annual GTC conference, introducing seven new chips designed for artificial intelligence workloads. The platform includes the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, Spectrum-6 Ethernet switch, and the integrated Groq 3 LPU inference accelerator.

CEO Jensen Huang described the platform as "a generational leap" that would enable "the greatest infrastructure buildout in history." The company claims the Vera Rubin platform delivers up to 10 times more inference throughput per watt and one-tenth the cost per token compared to its current Blackwell systems. Major AI companies including OpenAI, Anthropic, Meta, and Mistral AI have endorsed the platform, along with all major cloud providers.

The announcement was part of a broader product launch that included the Agent Toolkit, an open-source platform for building autonomous AI agents. Seventeen enterprise software companies, including Adobe, Salesforce, SAP, and ServiceNow, announced plans to integrate the toolkit into their platforms. The toolkit includes Nemotron models, OpenShell runtime for security, and optimization libraries designed to help AI agents operate autonomously within organizations.

Nvidia also introduced the DGX Station, a desktop supercomputer powered by the new GB300 Grace Blackwell Ultra Desktop Superchip. The system can run AI models with up to one trillion parameters and delivers 20 petaflops of compute performance with 748 gigabytes of memory. The machine is designed to enable local AI development and deployment without cloud infrastructure.

The company announced partnerships across multiple industries, including autonomous vehicles with BYD, Geely, and Nissan, and healthcare robotics with Johnson & Johnson MedTech and Medtronic. Pharmaceutical company Roche disclosed plans to deploy more than 3,500 Blackwell GPUs across its operations. Nvidia also revealed a space-focused version of the Vera Rubin platform for orbital data centers, claiming 25 times more AI compute performance than its current H100 GPU for space applications.

The announcements position Nvidia to supply AI computing infrastructure across scales from desktop to data center to space-based systems. All major cloud providers including Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure committed to offering the Vera Rubin platform, with more than 80 manufacturing partners building systems around the technology.