Nvidia CEO says his AI chips are improving faster than Moore’s Law

Nvidia CEO claims AI chips are improving faster than Moore's Law.

: Nvidia CEO Jensen Huang argues that his company's AI chips are progressing at a rate faster than Moore's Law, traditionally used to gauge computing advancement. Nvidia's latest data center superchip, the GB200 NVL72, is 30 to 40x faster for AI inference than its predecessors. This improvement is due to innovations across the architecture, chip, system, libraries, and algorithms. Huang emphasizes that improving performance will lower AI inference costs, making advanced AI models more accessible.

Nvidia CEO Jensen Huang claims that Nvidia AI chips are advancing at a rate faster than Moore's Law, a longstanding measure for computing progress. In an interview with TechCrunch, following a keynote at CES, he noted that the company's newest superchip is 30 to 40x more efficient for AI tasks than previous models.

Huang explains the significance of developing the architecture, chip, system, libraries, and algorithms in unison, which accelerates progress beyond Moore's Law rate. This advancement is particularly important as it can reduce the cost of inference, a crucial phase in AI model operation, making them more economically feasible.

Amid discussions about AI's stalled progress, Huang highlights three active AI scaling laws and ensures that Nvidia chips meet these needs efficiently. The latest advancements are crucial for companies like Google, OpenAI, and others that rely on Nvidia chips to enhance AI model capabilities and affordability.

AI chipsMoores LawNvidia