Agent AI (Agent-Based AI) has arrived! The next wave of AI is shifting from generative AI to agent-based AI.
On June 1st, 2026, at the GTC conference held in Taipei, China, the founder and CEO of NVIDIA, Huang Renxun, delivered a keynote speech. In this event broadcast to 70+ countries and regions around the world, Huang Renxun not only unveiled a series of new products, but also systematically expounded his core judgments on the current and future development of the AI industry.
At the product level, NVIDIA officially announced that its most ambitious Vera Rubin AI system has achieved full-scale production and has launched the Vera CPU specifically designed for the AI era. At the same time, the company released an open-source AI model named Nemotron 3 Ultra, as well as the RTX Spark super chip for personal computers running the Windows system, marking NVIDIA's official entry into the PC chip market. Additionally, NVIDIA has launched the DSX platform for building AI factories, the Alpamayo 2 inference model specifically for robot taxi services, and the Isaac GROOT development platform for humanoid robots.
Huang Renxun clearly declared in his speech: "Agent AI (Agent AI) has arrived." He pointed out that the next wave of AI is shifting from generative AI to agent AI, and its core feature is the ability to understand human intentions and autonomously invoke tools to complete tasks, which will completely change the way humans interact with machines. He demonstrated how AI Agents can automatically generate animations or CAD designs based on text instructions, and asserted: "This is how future computers will operate."
In response to the industry's concerns about the impact of AI on employment, Huang Renxun refuted by saying "This is completely nonsense." He used data from the GitHub platform as an example, pointing out that the usage of AI programming increased from 300 million times in 2023 to 1.4 billion times in the first few months of 2026, while the number of software engineers actually increased. He emphasized that AI did not reduce job opportunities; instead, it created new demands and ecosystems.
From a business value perspective, Huang Renxun put forward the viewpoint that "Token is an asset". He believes that Tokens have become the revenue unit for AI companies to make profits, driving a huge global demand for AI computing power. He described this demand as "rocketing upward" and humorously compared it to the trend chart of the Taiwan stock market.
Regarding the emerging business model of AI factories, Huang Renxun revealed their astonishing capital intensity. He disclosed that the initial cost for a 1GW-level AI factory could reach between 20 billion and 30 billion US dollars, and in the future, it might even reach 80 billion to 100 billion US dollars per GW. To this end, NVIDIA's DSX platform is designed to enable builders to complete the entire factory's design and verification in a digital simulator before investing any funds.
When discussing the software industry, Huang Renxun holds an optimistic attitude. He believes that the popularization of agent-based AI will not pose a threat to software companies; instead, it will bring huge development opportunities due to the need to call upon more tools. At the same time, he announced that NVIDIA and Microsoft have had in-depth cooperation over the past three years, "and are about to redefine personal computers". The AI PC equipped with the RTX Spark chip will be launched by manufacturers such as Dell and Lenovo this autumn, and at that time, the AI Agent will run directly on personal computers.
At the end of the speech, Huang Renxun summarized the evolution of NVIDIA's positioning: "NVIDIA has indeed become an infrastructure company." He stated that the company's goal is not only to provide GPUs or systems, but also to assist customers in building AI infrastructure that can generate the highest revenue and profits. He envisioned that the model of proxy AI will expand from computers to robots, satellites, cloud base stations and factories, profoundly changing the definition of various devices in the future.
