Nvidia CEO Huang Says New Vera Rubin Design Is in Full Production
Speakers
Jensen Huang
— CEO, NVIDIA
Summary
Jensen Huang's CES presentation argues that AI compute demand is exploding as models scale, inference becomes a reasoning process, and test-time compute rises. He says NVIDIA GPU demand is skyrocketing and that the industry's race to the frontier is fundamentally a computing problem. He also announces that GB300 is in full-scale manufacturing and the new Vera Rubin design is in full production for this year.
- AI computation requirements are rising rapidly.
- Model sizes are growing by about 10x every year.
- Inference now uses reasoning, reinforcement learning, and test-time scaling.
- Token generation from test-time scaling rises about 5x yearly.
- Cost of last-generation AI tokens falls about 10x yearly amid intense competition.
- NVIDIA GPU demand is described as skyrocketing.
- Nvidia emphasizes annual advancement of state-of-the-art computation.
- GB300 is in full-scale manufacturing and Vera Rubin is in full production.
Ideas
AI compute boom drives NVIDIA GPU demand
AI computation demand is skyrocketing because model sizes grow about 10x annually, inference has shifted to a reasoning/thinking process using reinforcement learning in post-training, and test-time scaling is increasing token generation by about 5x yearly. The intense race to the AI frontier makes faster compute critical, and Huang explicitly says demand for NVIDIA GPUs is skyrocketing.
Up Next
This Bloomberg Markets video, published January 05, 2026, features Jensen Huang discussing NVDA. 1 trade idea extracted by AI with direction and confidence scoring.
Speakers: Jensen Huang · Tickers: NVDA