Jensen Huang slams Wall Street: You misunderstood DeepSeek! Chip demand will surge 5–10 times

In an Axios interview, Huang Renxun directly pointed out that Wall Street has misunderstood China’s AI models twice, saying DeepSeek and Kimi will drive more compute demand; the chip industry needs to expand by 5–10 times, with no bubble risk within five years. He also blasted AI doomsday talk as nonsense and revealed that NVIDIA’s sales in China are nearly zero.
(Background: “Lin Shanglun, Esq. special article》We created the world’s strongest AI chips, yet the clerk typed herself into despair”)
(Background supplement: Anthropic no longer shows Claude’s thought-process summary to prevent Fable 5 from being distilled)

Table of contents

Toggle

• Wall Street’s double misjudgment: Why are DeepSeek and Kimi a positive?
• AI apocalyptic theory gets slapped down: Radiologists don’t decline but increase by 20%
• Chip expansion by 5–10 times: No bubble risk within five years
• Tokenomics: Smarter AI is worth more
• The era of trillion-agent: Every company needs AI employees

NVIDIA’s CEO, Huang Renxun, whose company has the highest global market cap, recently went all out in an Axios interview. He directly said Wall Street’s panic about China’s AI models DeepSeek and Kimi was a misread, stressing that good models only bring more chip demand; meanwhile, he also denounced AI doomsday as “nonsense” and offered a shocking forecast that the semiconductor industry needs to expand by 5–10 times. This 70-minute conversation almost covers every hot dispute in the AI industry.

Duration: 70 minutes

“Calling AI will destroy humanity is completely nonsense. Saying AI will eliminate half of American jobs is completely nonsense. All facts and evidence point in the opposite direction.”

“China’s AI researchers are more than all other places in the world combined. China will become extraordinary in this field—this is destined. Want to stop China? That’s a foolish idea, and it’s simply not doable.”

“The semiconductor industry needs to expand 5 to 10 times. Everything is in short supply today—chip shortages, memory shortages, shortages of land, power, and construction workers. This kind of shortage is actually a good thing. It gives us time to build infrastructure.”

“OpenAI and Anthropic will become the most successful IPOs in human history. One company is worth a trillion dollars after only a few years—how could that be? Because AI is useful, and if AI is useful, it can make money.”

Host Mike Allen opened by throwing out the most sensitive topic: An FT report said China is considering tightening export controls on AI models and semiconductors; Kimi, a China model, was released and Nvidia’s stock price promptly fell, with chip stocks down 18% in one month. What exactly did Wall Street get wrong?

Wall Street’s double misjudgment: Why are DeepSeek and Kimi a positive?

Huang Renxun’s answer was straightforward. He said the market got it wrong once when DeepSeek came out, and now it’s making the same mistake again. “A good model leads to more usage. More usage means you need to sell more Nvidia computers. You need to build more data centers and provide more services, and then the technology penetrates more industries.” He repeatedly emphasized this chain of logic: a good model brings good applications, good applications bring growth, and growth requires more compute.

When asked whether Kimi and other China models should be shut down, his answer was just as direct: “Of course you should use it. That’s smart.” He explained that after downloading a model, you can fine-tune, enhance, and add guardrails; the model runs inside so-called “harness” (frameworks), and the framework is in a “sandbox.” The sandbox is safe, with privacy protection, safety controls, and access controls. He compared it to operating systems: Linux is open source, reviewed and tested and hardened by millions of people worldwide—so we can trust it. Open-source AI models are the same logic.

He also pointed out an easily overlooked point: open models and closed models are not opposites. “The most likely people to upgrade to good models like Anthropic or OpenAI are those who are already using AI.” Free models lower the barrier to trying AI. Once users get started, they naturally want better services, and that’s when they pay closed-model vendors. So the more open models there are, the more opportunities there are for closed models.

When asked about NVIDIA’s sales in China, Huang Renxun gave an answer that many people didn’t expect: “Our sales in China today are about zero.” He said he has already told investors not to expect any revenue from China. If the Chinese government and market welcome them back, that would be “a great honor.” Until then, treat sales as zero.

This was the most explosive part of the entire conversation. The host asked: Are some of your tech-industry peers overhyping AI risks?

Huang Renxun responded without politeness. “It’s fine to warn people, and it’s even better to warn them with solutions, but fabricating facts is absolutely not allowed.” He then named several popular AI panic narratives: “Saying AI will end humanity is completely nonsense. Saying AI will wipe out half of American jobs is completely nonsense. All facts and evidence point in the opposite direction.”

He gave several specific examples to support his argument. The number of radiologists increased by about 20% because AI automated scanning and analysis, allowing doctors to see more patients; since there are too many people needing medical care, more radiologists are actually needed. The number of legal assistants increased by about 10% as well—same logic. Employment in manufacturing grew by about 50% over the past few years because AI data centers need to be built, and chips need to be produced.

“Intuition and common sense tell you that AI increases productivity. Increased productivity creates opportunities. Looking across history, technology makes society more efficient and creates more jobs rather than fewer. If that weren’t the case, the U.S. would have only 100,000 jobs right now.”

AI apocalyptic theory gets slapped down: Radiologists don’t decline but increase by 20%

He also turned his fire on AI leaders. “Doomsayers spend too much time theorizing sci-fi endings. Maybe that makes them seem smart.” When pressed on whether he was talking about the CEO of certain AI companies, he didn’t deny it; instead he said: “If you want the world to be wary of this incredible capability of the technology, that goal has already been achieved. We should spend our time on making the technology safe—that’s the responsibility of technology leaders.”

The host mentioned that Asia’s attitude toward AI is starkly different from America’s; Huang Renxun was surrounded by fans in Asia and asked for signatures. His explanation was thought-provoking: “Maybe it’s because the doomsayers spend too much time fabricating sci-fi endings.” In Asia, people embrace AI as a tool and opportunity rather than fearing it as a threat.

The host raised a pointed question: Every industrial revolution goes through bubbles. Where is the bubble risk in this generation?

Huang Renxun’s answer was cautious yet clear. “Bubbles will come one day, but not today. We’re at the very beginning of building.” He provided a time frame: unlikely within five years; depending on conditions between five and ten years. The reason is comprehensive constraints on the supply side. “The industry can build faster now, but chips aren’t enough, memory isn’t enough, land isn’t enough, power isn’t enough, and even construction workers aren’t enough. We are constrained in every direction and in every aspect.”

He said these constraints are a good thing. “These constraints hold the system back and give us enough time to build infrastructure.” Demand is strong, but converting demand into the capacity to build productive supercomputers is delayed because of all these physical limitations. That pushes the point where supply exceeds demand back.

More importantly, he offered his overall assessment of the semiconductor industry’s scale. “I believe the semiconductor industry needs to expand by 5 to 10 times.” Over what time span? “In the next ten years.” That means today’s chip industry is still far too small—nowhere near enough to support the construction demand for AI infrastructure.

He explained why this semiconductor cycle is different from the past. “This time is different because it’s not demand-driven, not seasonal, and not consumer-driven. It’s driven by industrial infrastructure.” Just like the world needs energy, the internet, roads, and railways, it now needs an AI layer of intelligent infrastructure built on top of all existing infrastructure. That layer needs chips.

On the risk of customers taking on debt to buy NVIDIA products, he said he’s not too worried. “These companies are excellent. They generate a lot of cash.” He specifically mentioned the startup of the AI profit flywheel: AI is useful, so AI can make money. Coding agents are extremely profitable—they do useful work in high-paying roles, and many companies are willing to pay hundreds of millions of dollars every year to enhance their coding capabilities. “The flywheel has already started.”

Chip expansion by 5–10 times: No bubble risk within five years

This was the most technical deep-dive part of the whole conversation. The host asked: What makes you confident that tokens will keep making more money?

Huang Renxun’s explanation was vivid. “Tokens are a type of embedding. They embed knowledge and intelligence. This number isn’t static, like pi. The intelligence encoded in this number becomes smarter over time.” As intelligence becomes smarter, it becomes more useful. The more useful it is, the more valuable it is, and then people are willing to pay more.

He compared the AI industry with the software industry of the past. In the past, software was “light-capital,” so software companies had high gross margins. But in the AI era, the software industry will become more “heavy-capital,” because producing modern software that generates intelligence needs machines like the supercomputer in front of him. Every industry will become more capital-intensive, but the payoff is incredible intelligence, productivity, and growth.

“We are laying the foundation, building infrastructure—this is the largest-scale industrial infrastructure buildout in human history.”

On the question of whether AI has already peaked, his answer was: “It can’t have peaked because AI penetration into society and industries has only just begun.” He also mentioned that in the past six months alone, risk capital and startups in the U.S. totaled $300 billion, all of which have created new jobs and new companies.

Huang Renxun’s assessment of Trump was unexpectedly specific. “He’s smart, he remembers everything. Oh my God, he actually understands numbers.” He said Trump is the only president who can remember NVIDIA chip model names like H20, H200, and Blackwell—and even knows the name of the next-generation product, Rubin.

He described their first meeting with Trump: “He said he wanted to restore the manufacturing capacity of the U.S., to reindustrialize the U.S., to have secure and resilient supply chains, and to bring semiconductor manufacturing back to the U.S.” He said the Fort Worth factory where they sat for the interview was the direct result of that conversation.

When asked what mistakes the government should avoid making, Huang Renxun rarely showed anxiety. “What I worry about is overregulation and overcorrection.” He said some companies want the government to help write regulations favorable to them. “I think we should compete in the open.” He dismissed the AI race as a “100-meter dash” with a comment he called “nonsense”: “Whoever runs to the finish line first wins forever? That’s nonsense. Ultimate victory is whether a society uses the technology, regardless of who invented it. We didn’t invent electricity or manufacturing, but the U.S. adopted them faster and with more passion—so the U.S. became what it is today.”

Tokenomics: Smarter AI is worth more

The host pressed further: If Trump called asking for equity in NVIDIA, what would you say? Huang Renxun answered cleverly: “No need. The U.S. already has equity in NVIDIA. Last year we paid $50k in taxes; this year we’ll pay more. We create a lot of jobs and taxes. And don’t forget: most Americans are in the stock market today. When the stock market goes up, everyone benefits.”

Regarding whether Anthropic’s Mythos model should be open to everyone (currently only open to specific institutions), Huang Renxun was very clear: “Of course it should be open to everyone.” He said it’s Anthropic’s responsibility to ensure technical safety, just like any software—if vulnerabilities are found, patch them as quickly as possible. He mentioned the jailbreak incident of Mythos, saying, “Everything’s fine. You and I are still chatting here.”

On the issue of open-model companies distilling closed models and then reselling them, his stance was more balanced: “It depends on the terms of service. If the service provider is unhappy, they should contact that company. We have a lot of traditional legal tools to handle this.” But he emphasized that learning from other sources is a basic characteristic of AI. “AI must learn from something. Even raw AI originally scraped all existing knowledge from the internet. Now AI-generated content is more than humans. In a few years, 99% of content on the internet will be AI-generated. You’re already constantly distilling other AI’s intelligence.”

The host asked: When will the ChatGPT moment for robots arrive?

Huang Renxun gave an unexpected answer: “The ChatGPT moment for robots has already arrived.” He explained that the ChatGPT moment wasn’t the moment AI became useful. When ChatGPT came out in 2022, it was just “interesting and surprising” to people. It took four more years to become truly useful. Robots are now in that “interesting and surprising” phase. You can tell a robot, “Put the apple in the drawer,” and it will infer a sequence of tasks, including opening the drawer first and then putting in the apple. “When you see a mechanical robot actually do these things for the first time, it opens up your imagination about the future of robots.”

As for when it will become useful, he gave a time frame of “three to four years,” and said “I won’t be surprised.”

On the era of agents, he painted a striking picture. Today, a billion people use computers, but most of the time computers are idle. In the future, everyone will be assisted by large numbers of agents, and these agents will use computers all day, every day. “We’ll have 100 billion, trillions of agents running all day, every day. Intelligent agents, not-so-intelligent agents, specialized agents, super agents—every kind of agent running all day, every day. So the number of computers we need will grow dramatically.”

The conversation ended back on Huang Renxun personally. The host mentioned that he founded NVIDIA at age 30; today the company is worth $5 trillion, but with only 50k employees. He said that in ten years there might only be 75k people too—“as small as possible.”

The era of trillion-agent: Every company needs AI employees

Huang Renxun defined the CEO’s job as “strategy.” “Strategy is to realize a future vision as efficiently as possible with limited resources.” He said he started doing this job at age 30, which makes him likely the longest-tenured CEO in technology history. “This is my craft. This is my skill.”

On the theory that “pain and grinding are the key to greatness,” he corrected it: it’s not related to any specific painful incident; it’s more like an ongoing state. “No great athlete becomes great by chance. It’s a large amount of practice when nobody’s watching, a large amount of failures, a large amount of pain and grinding.” He said this grinding improves craft, shapes character, and gives people confidence and resilience. “Like athletes say, when the pressure and tension are highest, time feels like it slows down. I feel the same. That comes from practice—doing the same thing again and again.”

When asked what he would tell himself if he had arrived in the U.S. at age nine, his answer was a passionate praise of America: “America is the greatest country in the world, period. Even if we have challenges and disagreements, those challenges and disagreements are exactly what make us great. Through open dialogue, through a vibrant system that allows freedom—freedom to speak, freedom to innovate, freedom to start businesses, freedom to collaborate.” He concluded: “This country is built by immigrants, and it will still need remarkable immigrants in the future.”

On his habit of not wearing a watch, his explanation was very Huang Renxun: “Because this is the most important time right now. I refuse to let my schedule manage my life, and I refuse to let a watch manage my life. If I’m late, someone will tell me. Until then, I’m here 100% of the time.”

• Conflict of interest: Huang Renxun holds about 3.5% of NVIDIA shares (about 860 million shares), with personal wealth of about $181 billion (Bloomberg data for June 2026). The content of this issue involves topics such as AI industry expansion, growth in chip demand, and reduced regulation, all of which are directly related to his personal financial interests. His argument that “good models bring more chip demand” is essentially saying China’s AI models are a positive for NVIDIA; readers need to judge for themselves.
• About “China sales = ‘approximately zero’”: Huang Renxun claims NVIDIA’s sales in China are “about zero.” According to publicly available financial reports, U.S. export controls have indeed restricted NVIDIA’s high-end chip sales to extremely low levels in China, but “zero” is Huang Renxun’s phrasing; the actual figures need to be verified via financial reports.
• About data on AI creating jobs: The figures that radiologists increase by 20%, legal assistants increase by 10%, and manufacturing increases by 50% are cited as spoken by Huang Renxun and are not labeled with sources in the conversation. Readers can cross-verify independently.