Jen-Hsun Huang: It is expected to take 10 to 20 years for the U.S. chip supply chain to become independent, and AI is expected to surpass humans within 5 years

November 30 news, Nvidia CEO Jensen Huang recently attended the DealBook Summit hosted by the New York Times and discussed the development trends of artificial intelligence, the impact of the U.S. semiconductor export control policy, etc. He also pointed out that the U.S. chip supply chain is independent It will take at least another 10-20 years to achieve this.

It will take 10-20 years for the U.S. chip supply chain to become independent

Huang Renxun said in the interview that there are 35,000 components in Nvidia's system from suppliers around the world, 80% of which are provided by TSMC and its supply chain, which are located all over the world. "It will take the United States 10 to 20 years to achieve supply chain independence. In one or two years, this is not something that can be achieved, and supply chain independence will be very challenging."

However, Huang Renxun also emphasized, "We should try. We should work hard. I should definitely go on."

Currently Nvidia also has plans to manufacture chips in the United States. According to Core Intelligence, Nvidia Chief Financial Officer Colette Kress pointed out this week that TSMC’s Arizona factory may be the “third foundry” for Nvidia GPUs. Prior to this, Huang Renxun also said that Intel would also be considered as a chip foundry supplier in the future.

The impact of new US semiconductor regulations

Regarding the impact of the new U.S. restrictions on Chinese semiconductors on Nvidia, Huang Renxun said that U.S. national security and national competitiveness are naturally issues that need to be considered, but “we are a company born for business, and we will try our best to cooperate with Business for all”.

Mainland China is still the largest chip market, and Nvidia hopes to continue to be committed to the development of the Chinese market. Nvidia is also currently developing special chips for the Chinese market that will not violate U.S. export controls.

Huang Renxun has previously said: "We must develop new chips that comply with export control regulations. Once the requirements are met, we will return to China to sell. We will try our best to do business with everyone."

The restrictions on U.S. semiconductor exports to China have further stimulated China’s determination to develop its own independent semiconductor industry. Huang Renxun also said that dozens of companies in mainland China are currently developing technologies that can compete with Nvidia's products, which is also detrimental to Nvidia's development in the Chinese market.

Cooperation with OpenAI

However, Huang Renxun emphasized that he is not worried about intensifying competition. He said Nvidia has been ahead for a decade since it started producing the first chips for supercomputers in 2012. Supercomputers based on Nvidia chips due to be delivered in 2017 will change the way computers work, enabling them to generate sentences, images and videos. Here I have to mention the cooperation between NVIDIA and OpenAI.

In the interview that day, Huang Renxun also recalled the scene when he delivered "the world's first AI supercomputer" to OpenAI. At that time, Elon Musk, the co-founder of OpenAI, had not yet left the company. Musk heard Huang’s introduction to the device at a conference in 2017.

Huang Renxun recalled: "Musk heard my introduction and said, 'I want one too' and told me about OpenAI. On that day, I delivered the world's first artificial intelligence to OpenAI supercomputer."

Mentioning the recent reorganization of OpenAI’s board of directors and the dismissal of CEO Sam Altman, which has attracted the attention of the entire network, Huang Renxun said that he hopes that things will calm down as soon as possible: “I am very happy that they have reached an agreement. I also hope they can reach consensus. This is a very great team. This also reminds us of the importance of corporate governance. NVIDIA has experienced a lot of adversity after it was founded 30 years ago. Building a company, who knows what will happen.

AI may surpass human intelligence within 5 years

Regarding the current artificial intelligence craze, Huang Renxun believes that if artificial intelligence (AGI) is defined as a computer that can complete tests in a manner that is "quite competitive" with human intelligence, then "in the next five years, artificial intelligence will be able to complete these tests" Test.” Until then, he predicts companies will develop custom AI capabilities to achieve specific goals, such as drug research or designing semiconductors.

Huang Renxun also said: "We realize that deep learning and artificial intelligence are not a chip problem. This is a reinvention of computing. You cannot solve this new way of computing just by designing chips. All aspects of computers are happening Fundamental changes.”

"NVIDIA's core business already relies on artificial intelligence. Without artificial intelligence, our chips today would not be possible." Huang Renxun further explained that NVIDIA's H100 chip was designed with the help of a variety of AI, "Without With AI, you can’t write software, you can’t design chips, everything is impossible.”

Editor: Xinzhixun-Rurounijian

The above is the detailed content of Jen-Hsun Huang: It is expected to take 10 to 20 years for the U.S. chip supply chain to become independent, and AI is expected to surpass humans within 5 years. For more information, please follow other related articles on the PHP Chinese website!