NVIDIA CEO Jensen Huang opened CES 2026 at the Fontainebleau Las Vegas by declaring a total modernization of the world’s $10 trillion legacy computing footprint. The keynote shifted NVIDIA’s focus from just “selling chips” to delivering fully integrated AI factories.
The Rubin Platform: 6 Chips, 1 Supercomputer
NVIDIA’s flagship announcement was the Rubin platform, succeeding the record-breaking Blackwell architecture. Rubin is the industry’s first “extreme codesigned” system, built to slash the cost of generating AI tokens by up to 90%.
Key Technical Specifications:
- Rubin GPU: Features HBM4 memory and delivers 50 PFLOPS of NVFP4 inference performance (a 5x leap over Blackwell).
- Vera CPU: An 88-core custom Arm-based processor designed specifically for agentic inference and data movement.
- Networking: Integrated NVLink 6 (6 TB/s) and BlueField-4 DPUs for seamless data-center-scale workloads.
Also Read: Marvell to Acquire XConn Technologies, Expanding Leadership in AI Data Center Connectivity
Physical AI: From Simulation to the Mercedes CLA
A major theme of 2026 is the expansion of AI into the physical world. NVIDIA introduced Cosmos, a world foundation model designed for simulation, and Alpamayo, a suite of reasoning vision-language-action (VLA) models.
Key Milestone: The first passenger vehicle powered by the Alpamayo/NVIDIA DRIVE full-stack platform will debut on U.S. roads this year in the new Mercedes-Benz CLA. Unlike previous systems, Alpamayo uses “chain-of-thought” reasoning to explain its driving decisions in real-time.
Gaming: DLSS 4.5 and 6X Frame Generation
For the consumer market, NVIDIA unveiled DLSS 4.5, featuring a second-generation transformer model.
- Dynamic Multi Frame Generation: An “automatic transmission” for GPUs that shifts between frame multipliers to hit 240+ FPS.
- 6X Mode: Allows for path-traced 4K gaming on next-gen RTX hardware with unprecedented smoothness.
