Hwang Jensen, NVIDIA CEO, speaks at an event. /Courtesy of News1

Jensen Huang, Nvidia's chief executive officer (CEO), said that if the Trump administration's H-1B visa policy had been in place when he came to the United States, his family would not have been able to come to the country.

Huang said on the 8th (local time) in an interview on CNBC's program Squawk Box that "if the current Trump administration's policy had applied back then, our family's immigration to the United States would have been impossible." Born in Taiwan, Huang moved to Thailand and immigrated to the United States with his older brother at age 9, with his parents joining about two years later.

President Donald Trump last month announced that when applying for H-1B visas, which target foreign skilled workers, employers must pay a $100,000 (about 142.1 million won) fee per visa. That is a 100-fold increase from the previous level.

Huang said, "Our family would not have been able to afford $100,000," adding, "So the very opportunity for my family and me to come here would have disappeared." He continued, "Immigration is the foundation of the 'American dream,' which means the ideal that anyone can build a better future through effort and talent," and added, "Our family came to the United States for more opportunities and to enjoy the benefits of this amazing country."

Nvidia is currently sponsoring 1,400 visas. He said, "We will continue to cover employees' H-1B visa expense," and expressed hope that "there will be some improvements to the policy so that the chance for serendipity to happen still exists." He added that while Trump's new immigration policy could block immigration by families like his, "the United States will nevertheless still be able to attract the world's best talent."

Huang also said he was "surprised" that AMD recently signed a multibillion-dollar partnership with OpenAI and agreed to provide 10% of its equity to OpenAI. He said, "(Providing equity) is imaginative, unique, and surprising," adding, "Especially considering how much AMD was counting on its next-generation products." He continued, "It's surprising to give away 10% of the company before making the product," and added, "In my view, it's clever."

Meanwhile, AMD announced on the 6th that it had signed a multi-year contract to supply OpenAI with AI chips worth tens of billions of dollars annually. The graphics processing units (GPUs) supplied by AMD amount to 6 gigawatts (GW; 1 GW is the generating capacity of one nuclear power plant) when converted to power. In particular, AMD also granted OpenAI the option to acquire up to 10% of its equity.

Nvidia earlier announced on the 22nd of last month that it plans to invest up to $100 billion (about 142.17 trillion won) in OpenAI over the next 10 years. OpenAI agreed to build and operate Nvidia systems with 10 GW of power consumption, equivalent to 4 million to 5 million GPUs. Huang said, "This investment structure is very different from AMD's transaction," and explained, "Nvidia can sell products directly to OpenAI."

However, after Nvidia's OpenAI investment announcement, concerns were also raised about round-tripping in AI infrastructure transactions. The question was whether OpenAI would buy Nvidia chips with the money Nvidia invested. Asked how OpenAI could raise transaction funds with Nvidia, he said, "They don't have that money yet," adding, "Revenue is growing exponentially, but funds will have to be raised through sales growth, stock issuance (equity), or liability." He continued, "When that time comes, we will be given the opportunity to invest alongside other investors," and added, "What we regret about our past investment in OpenAI is that we couldn't invest more."

Huang also said that computing demand has surged over the past six months. Asked what question investors ask most, he said, "As AI models evolve from simple question-and-answer to performing complex reasoning, computing demand has surged this year, especially over the past six months." He added, "AI inference models use an enormous amount of computing power, but their outputs are so strong that demand is exploding."

Asked which country is ahead in the AI race, he answered, "Right now, the United States is not that far ahead of China." He said, "China is building out the power infrastructure to support AI much faster," adding, "In terms of energy, China is far ahead."

He emphasized that to meet the surging demand in the AI industry, companies must build their own power generation infrastructure without relying on the grid. He said, "If you rely only on the grid, the pace is too slow," and predicted, "Data centers will start with self-generation using natural gas and, going forward, will have to consider nuclear power." He added, "We have to invest in energy production in every way possible," and emphasized, "It can be much faster for data centers to produce power on their own than to connect to the grid, and we should move in that direction."

※ This article has been translated by AI. Share your feedback here.