Nvidia CEO Jensen Huang stated that his company’s chips outperform the pace set by Moore’s Law . Huang made this statement at the CES conference in Las Vegas.
Moore’s Law, formulated in 1965, predicted that the number of transistors on chips would double every year. However, in recent years this process has slowed, leading some analysts to claim that the law no longer holds. Huang, however, claims that Nvidia’s chips, particularly the new superchip for data centers, outperform the previous generation by 30 times. According to him, Nvidia’s chip performance today is 1,000 times higher than it was 10 years ago, indicating a rapid pace of development that is unlikely to slow down anytime soon.
Previously, laboratories at Google, OpenAI, and Anthropic, which use Nvidia chips to train their models, reported challenges with further scaling of AI. Improvements in Nvidia’s chips could contribute to the continued advancement of AI model capabilities.
Regarding cost, Huang stated that increased chip performance will lead to lower computing expenses. He emphasized that the new GB200 NVL72 superchip, which he demonstrated on stage, is 30–40 times faster than previous models. This means that models requiring significant computing resources for testing will become cheaper over time.
