NVIDIA CEO: AI to Follow Hyper Moore's Law

NVIDIA CEO Jensen Huang predicts that computational power will experience exponential growth over the next decade, leading to what he describes as "hyper Moore's Law," which will drive new innovations and capabilities in artificial intelligence (AI).

Speaking on the podcast No Priors on November 7, Huang stated, "In the coming decade, we expect to double performance every year or two." He added, "If Moore's Law traditionally saw performance doubling every few years, I wouldn't be surprised if we're now on a trajectory resembling some kind of hyper Moore's Law."

Huang explained that if performance increases every year or two, the cumulative effect will be "remarkably substantial." Moore's Law, originally predicted by Intel co-founder Gordon E. Moore, suggested that the number of transistors on a chip would double approximately every 24 months, driving performance gains and cost efficiencies over time.

Last year, NVIDIA announced plans to accelerate the release cycle of AI data center products, reducing it from two years to one. According to Huang, this increase in performance can significantly reduce computational costs for businesses. He emphasized that these improvements are not only due to advances in AI chips but also rely on overall system enhancements, including software, networking, algorithms, and hardware advancements.