T
TubeFolio
Back to Dashboard

NVIDIA CEO Jensen Huang Leaves Everyone SPEECHLESS (CES Supercut)

Ticker Symbol: YOUJan 6, 2026

Summary

Alex presents a comprehensive vision for the future of AI infrastructure, centered on the transition from the Blackwell architecture to the newly announced Vera Rubin platform. He emphasizes that AI models are growing by an order of magnitude every year, a pace that traditional semiconductor scaling can no longer match. To solve this, Alex introduces 'extreme co-design,' where every component—from the CPU and GPU to the networking switches and liquid cooling systems—is engineered simultaneously to maximize throughput and efficiency.

The thesis focuses on the Vera CPU and Rubin GPU as the new benchmarks for the industry. Alex highlights that the Rubin GPU delivers five times the floating-point performance of Blackwell with only a 1.6x increase in transistors, achieved through innovations like the NVFP4 Tensor Core. Furthermore, he discusses the critical role of networking, asserting that NVIDIA is now the world’s largest networking company due to the success of Spectrum-X and the new BlueField-4 data processing units, which allow for massive memory expansion and faster data handling within the rack.

NVDA (NVIDIA): Alex describes the company's shift from a chip manufacturer to a full-stack AI systems provider. He explains that the Vera Rubin system can train a 10-trillion parameter model in one-fourth the time of previous systems. He also highlights the energy efficiency of the new architecture, noting that it uses 45°C liquid cooling to eliminate the need for water chillers, potentially saving 6% of global data center power.

Mentioned Stocks

NVDA
Sentiment: BUYAction: RECOMMENDED

Reasoning: Alex showcases the Vera Rubin platform as a revolutionary leap in AI computing, delivering 5x the performance of the previous Blackwell generation. He highlights NVIDIA's dominance in networking with Spectrum-X and the massive efficiency gains from liquid cooling and co-designed hardware, which drastically reduce the cost and time required to train frontier AI models.

Loading chart...