NVIDIA CEO Jen-Hsun Huang was a guest on Dwarkesh Patel's podcast and talked about the company's strategy, ASIC competition, the Chinese market, and supply chain moats, among other topics. When asked what NVIDIA would do if the deep learning revolution never happened, Huang gave an unexpected answer.
"Accelerate computing. It's the same thing we have been doing." Huang Renxun said,NVIDIA's core concept has always been to accelerate computing, combining the GPU and CUDA architecture with the CPU, and executing codes and algorithms on the GPU, thereby achieving 100 to 200 times performance improvement.
Huang Renxun pointed out that general-purpose computing is not ideal in many scenarios, and the large-scale parallel capabilities of GPU can cover a wide range of fields such as engineering, science, physics, data processing, computer graphics, and image generation. "Even if AI didn't exist today, NVIDIA would be very, very big."
However, he also admitted that thinking about a world without AI makes him "sad."
Huang Renxun also talked about the competitive pressure brought by ASIC. With the rise of customized chips such as Google TPU and Amazon Trainium, NVIDIA faces challenges from customer self-developed chips in the field of AI training and reasoning.
But he emphasized that NVIDIA's moat lies not only in the chip itself, but also in the CUDA ecosystem and global infrastructure.
