NVIDIA CEO Jensen Huang Leaves Everyone SPEECHLESS (Supercut)

Quick Overview

NVIDIA CEO Jensen Huang announces an accelerated annual rhythm for full-stack AI chip design, introducing the Blackwell NVL72 successor, Vera, and the next-generation networking/computing platforms like NVLink and the Omniverse digital twin platform, emphasizing the shift from CPU-centric to GPU-accelerated computing as Moore's Law slows for general-purpose chips.

Key Points: NVIDIA commits to an annual rhythm for full-stack AI chip design, launching Blackwell NVL72 and Vera in successive years. The company introduces NVLink, a Quantum-GPU interconnect, enabling the simulation of nature directly on GPU supercomputers. NVIDIA leads all open-source contributions across AI models, data, and libraries, citing dominance in benchmarks like LLM Safety Leaderboard and OpenASR. The shift in computing is characterized by a move from general-purpose CPU scaling to GPU-accelerated computing, with AI spend overtaking general-purpose spend by 2027. NVIDIA partners with global leaders like Siemens, FANUC, and Figure to build Digital Twin platforms, demonstrated via a factory simulation and a humanoid robot learning physics entirely in simulation. NVIDIA announced a partnership with Nokia to build AI-Native 6G on the new NVIDIA ARC Aerial RAN Computer. The company also announced DRIVE Hyperion, an essential computing platform for AVs, featuring the DRIVE AGX ECU, and highlighted Uber's adoption.

Context: NVIDIA CEO Jensen Huang delivers key announcements during a presentation, focusing heavily on the acceleration of AI hardware roadmaps (Blackwell, Vera), the convergence of classical and quantum computing via NVLink, the dominance of NVIDIA's open-source ecosystem, and the critical role of Digital Twins and AI in industrial and robotics applications. The presentation contrasts slowing general-purpose CPU performance gains with the exponential growth seen in GPU-accelerated computing.

Detailed Analysis

Jensen Huang detailed NVIDIA's aggressive, annual full-stack AI product cadence, announcing the Blackwell NVL72 successor, Vera, and emphasizing the shift from relying on traditional CPU scaling (which is slowing down) to GPU-accelerated computing, projecting AI spend to surpass general-purpose computing spend by 2027. Key hardware announcements included the NVIDIA NVLink Quantum-GPU interconnect, enabling direct simulation of quantum computers on GPU supercomputers, and the DRIVE AGX ECU for autonomous vehicles, noting major automakers like Lucid, Mercedes-Benz, and Stellantis are adopting DRIVE Hyperion. Huang also showcased NVIDIA's leadership in the open-source AI community, leading on numerous leaderboards (e.g., LLM Safety, OpenASR), and revealed partnerships with companies like Nokia for AI-Native 6G development using the Aerial RAN Computer. Furthermore, collaborations with industrial leaders (Siemens, FANUC) for Omniverse Digital Twins were highlighted, demonstrating complex factory simulations and training humanoid robots entirely in simulation using physics-accurate modeling, underscoring NVIDIA's comprehensive platform approach across AI, quantum, robotics, and enterprise.