NVIDIA’s Jensen Huang Declares Moore’s Law Dead, Introduces Huang’s Law; Performance Expected to “Double” in Just One Quarter

NVIDIA has set a new benchmark in artificial intelligence (AI) performance, achieving levels previously thought unattainable. According to CEO Jensen Huang, the significance of Moore’s Law in shaping technological advancements has diminished considerably in today’s landscape.

NVIDIA Challenges Moore’s Law, Emphasizes Native Technologies in AI Advancement

For context, Moore’s Law is essentially an observation that has historically indicated a correlation between the reduction of chip sizes (node shrinking) and the increase in computational performance, typically predicting advancements every two years, with smaller incremental upgrades anticipated annually. However, in the rapidly evolving realm of AI, NVIDIA reveals new and sophisticated architectures approximately every six months, demonstrating that the company leverages current industry trends to maintain their competitive edge. Jensen asserts that the relevance of Moore’s Law in contemporary times is minimal, declaring that “the sky is the limit”when it comes to innovation.

During a keynote presentation at Computex, hosted by Foxconn, Huang addressed concerns regarding the future of AI chip development. When questioned about a potential slowdown, he pointed to the increasingly significant physical constraints that have rendered Moore’s Law less effective. Specifically, he highlighted that the reduction in node sizes no longer correlates with performance enhancements comparable to the breakthroughs NVIDIA has achieved through alternative strategies.

One pivotal change in this landscape is attributed to advanced packaging technologies, notably the CoWoS (Chip on Wafer on Substrate) technique pioneered by TSMC, which has significantly influenced performance metrics. Additionally, NVIDIA’s proprietary NVLink interconnect technology facilitates enhanced chip performance by enabling the connection of multiple systems in rack-scale configurations. This innovation empowers NVIDIA to provide computational capabilities that were once deemed unattainable.

NVIDIA’s momentum shows no signs of slowing. Huang indicated that the company might adapt its product roadmap to a quarterly release schedule, a feat that many experts would previously consider implausible. Currently, NVIDIA operates on a six-month cycle for product launches, with recent releases like Vera Rubin and Blackwell Ultra illustrating this rapid pace. Clearly, NVIDIA is committed to sustaining interest and innovation in the AI sector.

Source & Images