By Wen Tsui
SAN JOSE, the United States, March 18 (Xinhua) -- Nvidia CEO Jensen Huang said the company expects at least 1 trillion U.S. dollars in demand and orders for its Blackwell and Vera Rubin systems through 2027, using this forecast to argue that artificial intelligence (AI) is entering a new phase of commercial growth.
The keynote, which Xinhua attended, took place at Nvidia's GTC 2026 on Monday at the SAP Center in San Jose, California. Nvidia reported that the company's annual developer conference attracted over 30,000 attendees from more than 190 countries and regions.
Huang explained that the next phase of AI will be driven less by training ever larger models and more by inference, the real-time use of trained models to answer questions, generate content and perform tasks. As the cost of generating AI output decreases and response times improve, Huang said AI services will become more widely used across both business operations and daily life.
Nvidia is using the event, being held from March 16 to 19, to introduce a series of new products aimed at that shift, including the Vera Rubin system, the Vera central processing unit (CPU), new Groq 3 LPX processors designed to speed up AI responses, Dynamo software for managing AI workloads, and NeMoClaw, a software layer for AI agents.
These announcements signal Nvidia's ambition to position itself as more than a chipmaker. Huang outlined the company's broader strategy to provide the systems needed to build and run AI services, ranging from large-scale data centers to software agents designed to assist individual users and businesses.
For audiences outside the tech sector, the keynote's significance lay not in a single new chip or software tool, but in Nvidia's broader vision for AI: shaping the next stage of systems that can operate continuously in the background, assist with digital tasks, and gradually redefine productivity.
Throughout his keynote, Huang said that AI is at an "inflection point of inference." He noted that AI is increasingly capable of doing productive work and that as its utility grows, so too will its demand for computational power, creating a cycle where usage and revenue rise in tandem.
However, the keynote's most significant social claim revolved around OpenClaw and personal AI agents. Huang explained that OpenClaw enables the creation of personal agents, describing it as the operating system for agent-based computing, and argued that every company now needs an OpenClaw strategy, calling it "the new computer."
Huang framed this idea within a broader historical context, asserting that computing has been reinvented multiple times -- first with the personal computer, then with the Internet, and later with the mobile cloud. He argued that AI represents yet another platform shift, one that could give rise to a new generation of major companies and fundamentally transform how computing is used.
This vision was echoed in the GTC Keynote Pregame, where speakers described agentic AI as a move beyond chatbots toward systems that can reason, plan, use tools, and take action. OpenClaw creator Peter Steinberger suggested that such systems could take over repetitive aspects of knowledge work, allowing people to focus on higher-order thinking and decision-making.
Nvidia also introduced NeMoClaw for the OpenClaw community, enhancing it with privacy controls, security policies, and safety guardrails to make always-on agents more secure and reliable. The move underscored one of Huang's central points: As AI becomes more capable and autonomous, trust, privacy, and control will become increasingly critical.
The success of Huang's vision will depend on lower operating costs, sufficient computing capacity, and whether users find AI agents useful and trustworthy. At GTC 2026, he emphasized that Nvidia is no longer just selling faster computing -- it is shaping a future where AI agents become integral to work and daily life. ■