At the recent CES event, NVIDIA CEO Jensen Huang stated that the performance improvement of the company's AI chips has surpassed the historical standards of Moore's Law.
Moore's Law, proposed by Intel co-founder Gordon Moore in 1965, predicts that the number of transistors on computer chips would approximately double each year, thereby doubling chip performance as well. However, in recent years, the progress of Moore's Law has significantly slowed down.
Image source note: Image generated by AI, image authorized by service provider Midjourney
Jensen Huang pointed out that NVIDIA's latest data center super chip runs AI inference workloads at over 30 times the speed of the previous generation. He stated, "We can simultaneously build architecture, chips, systems, libraries, and algorithms. If we can do that, we can surpass Moore's Law because we can innovate across the entire tech stack."
This statement is particularly significant amid widespread skepticism about whether AI progress has stalled. Currently, leading AI labs such as Google, OpenAI, and Anthropic are using NVIDIA's AI chips to train and run AI models, so advancements in these chips will directly impact the capabilities of AI models.
Huang also mentioned that there are currently three active AI scaling laws: pre-training, post-training, and computation during testing. He emphasized that Moore's Law is so important in computing history because it has driven down computing costs, and improvements in performance during inference will also lead to lower inference costs.
Although some are concerned whether NVIDIA's expensive chips can maintain a leading position in inference, Huang stated that the latest GB200NVL72 chip is 30 to 40 times faster than the H100 chip on inference workloads, making AI inference models more affordable.
Huang stressed that enhancing computing power is a direct and effective way to address the issues of performance and cost affordability during inference. He anticipates that as computing technology continues to advance, the costs of AI models will keep decreasing, although some models from companies like OpenAI currently have high operating costs.
Huang stated that today's AI chips are 1000 times more advanced than those from ten years ago, and this pace of advancement far exceeds Moore's Law, which he believes will not stop anytime soon.
Key Points:
🌟 NVIDIA CEO Jensen Huang stated that the performance improvement of the company's AI chips has surpassed Moore's Law.
⚡ The latest GB200NVL72 chip is 30 to 40 times faster on AI inference workloads than its predecessor.
📉 Huang predicts that as computing power improves, the operational costs of AI models will gradually decrease.