Nvidia CEO Jensen Huang says his company’s workforce will be dominated by artificial intelligence (AI) agents, with the number of human employees expected to significantly outnumber digital workers.
Speaking to reporters during a wide-ranging media Q&A on the sidelines of GTC 2026 in San Jose this week, Huang shared his vision for the future of enterprise operations and how Nvidia is navigating geopolitical tightropes such as the U.S.-China conflict and the global semiconductor supply chain.
While much of the excitement at this year’s GTC focused on NVIDIA’s expected USD 1 trillion in orders for its Blackwell and Vera Rubin chips by the end of 2027, Huang used the session to share his thoughts on where the company itself will be in 10 years.
“In 10 years, we hope to have 75,000 employees. We want to be as small as possible and grow as needed,” he said. “These 75,000 employees will be working with 7.5 million agents, who will be working around the clock…We’re going to solve some truly incredible problems.”
Rather than putting human engineers out of work, Huang argued that increased adoption of agent AI will only increase human workloads.
“It used to be that you’d create a product spec, and then your team would start working on it for a month. The next month, you’re working on something else. Life is pretty casual. Now that a month is 30 minutes, you’re always on the critical path. Thanks to AI, we can do things very quickly, and we’ll end up doing more.”
Legacy software defense
With enthusiasm for tools like Anthropic’s Claude Cowork and OpenClaw, there were concerns that AI agents would make traditional enterprise software platforms obsolete.
Huang challenged that premise, citing electronic design automation (EDA) software suppliers such as Cadence and Synopsys as examples. He argued that AI agents do not use stochastic generation to “manifest transistors from scratch.” Instead, they will serve as power users of existing enterprise software, fundamentally changing traditional software business models where growth is limited by the number of human users.
“Agent engineers are going to be using the same tools that we are using, because once they’re done using the tools, they need to turn it back into structured data that I can understand,” Huang explained.
“Will SQL go away because we have agents? No, SQL is where the ground truth of the business is stored. Because we have agents, the number of tools we have to license will probably explode, not the other way around.”
Because engineering and enterprise work requires accurate and deterministic results and cannot afford to be probabilistic, AI agents will have to rely heavily on legacy software to validate and structure their work, he said.
China’s orders and Taiwan’s dependence
With Asian markets watching the geopolitical tug-of-war over semiconductor dominance, Huang provided an important update on Nvidia’s operations in China amid continued U.S. export restrictions. The company has begun operations in China within the Trump administration’s trade framework.
“Will SQL go away because we have agents? No, SQL is where the ground truth of the business is stored. Because we have agents, the number of tools we have to license will probably explode, not the other way around.”
Jensen Huang, NVIDIA
“We have a license for the H200,” Huang said, referring to Nvidia’s previous generation Hopper graphics chips. “We have received orders from many customers in China and are currently restarting production… and the supply chain is also revitalizing,” he revealed.
He added that the administration wants to maintain the United States’ leadership in technology, but also recognizes the economic need for global commerce. “President Trump’s intention is for the United States to gain a leadership position and have access to NVIDIA’s best technology. However, President Trump wants the United States to compete around the world and not unnecessarily concede those markets.”
But Hwang poured cold water on the idea that the United States could quickly cut its semiconductor supply chain from Asia. In response to a question about the U.S. Department of Commerce’s goal of moving 40% of TSMC’s chip production to the U.S. mainland, Huang suggested that goal is unlikely in the short term.
“TSMC is doing its best to set up factories and supply chains in Arizona and across the United States. But as you know, demand is increasing very quickly,” he said. “New plants and factories are being built in the U.S., but the overall demand around the world is increasing very significantly. I think it will be very difficult to reach 40%.”
Industrial breakthroughs in Asia and Europe
During the session, Huang also made a pitch to manufacturing industries in Asia and Europe, offering a roadmap for major developed countries that missed out on the Internet software boom to regain global technological leadership.
Mr. Huang pointed out that countries such as Japan and Germany, historical leaders in the physical engineering field of mechatronics, were left behind in the IT revolution, and the United States came to dominate because the “ship before repair” culture in software development clashed with the safety-first principle of manufacturing.
But that all changes with the introduction of AI agents that can write their own code by following natural language prompts. “With OpenClaw, you don’t have to program it; you just tell it what to do,” Huang says.
“This is an opportunity to let bygones be bygones. It doesn’t matter anymore, because, as you know, software programmers don’t have to program anymore,” he added. “For Germany and Japan, this should be the best news in the world. If you dive straight into AI and inject AI technology with the genius of the mechatronics industry, suddenly you have a robotics industry.”
Self-driving cars and robotics
Huang defended Nvidia’s auto business, which currently accounts for only a small portion of its total revenue, saying the true scale of its business is hidden by the fact that global automakers not only buy chips for their vehicles, but also large clusters of Nvidia servers to train their AI models in data centers.
He also boldly declared that self-driving cars are a solved problem. “The rest is just engineering improvements,” he said, noting that if the industry achieves 1 trillion self-driving miles per day, it will be a multitrillion-dollar business.
Regarding robotics, Huang said he believes reasoning systems will rapidly advance robotics, as AI can now reason and break down complex physical environments into subtasks. “Once we see the evidence of the technology, it will take less than five years to improve,” he said. “I think we’re going to see some very good robots.”
