Jensen Huang’s keynote argues that AI has crossed from flashy demos into a profitable, production-ready computing shift: “useful AI has arrived.” He says agentic AI is now driving token demand, changing the role of CPUs and GPUs, and requiring a new end-to-end infrastructure stack from cloud to PC to robots. The presentation centers on Vera Rubin systems, Vera CPUs, NVIDIA’s enterprise agent toolkit, open models like Nemotron and Cosmos, and new PC/robotics platforms built with Microsoft, MediaTek, Cadence, and others.
Watch on YouTube ›Get the market thesis, key claims, assets, contradictions, and follow-up questions from any financial video — then unlock a version personalized to your portfolio, watchlist, and favorite speakers.
This keynote is essentially NVIDIA’s roadmap presentation for the “agentic age.” Huang’s core thesis is that AI has moved beyond generative novelty into a phase where agents can do useful work, generate economic value, and justify massive infrastructure spending. He frames this as a shift from AI as an output generator to AI as a profit generator: tokens are now “profitable units,” compute is revenue, and the bottleneck has become the ability to build enough AI factories to satisfy demand. He spends a large portion of the talk explaining the new agentic computing model. In his framing, an agent is not just a model; it is a model plus a harness, tools, memory, and a runtime. The harness orchestrates reasoning, planning, tool use, security, and memory management. That architecture, he argues, is fundamentally different from classic application software and creates demand for a new stack. …
Near term, the setup is bullish on narrative momentum and ecosystem follow-through, but crowded expectations make the main risk a miss in ramp timing, customer adoption, or benchmark credibility.
Over the next few months, the key question is whether agentic demand turns into visible bookings for Vera Rubin, Vera CPUs, and enterprise software adoption; if that happens, the platform story strengthens materially.
Structurally, Huang is arguing that AI agents create a new computing regime where NVIDIA can own the infrastructure layer across cloud, PC, robotics, and industrial systems; the thesis only weakens if agents fail to become durable economic workloads.
Useful AI has arrived and agentic AI is now the next major wave.
Core framing of the keynote and the basis for the rest of the product stack.
AI is increasing software developer output dramatically rather than eliminating software jobs.
He points to GitHub commit activity and argues more productivity will lead to more hiring.
Tokens are now profitable units, so AI companies will demand far more compute and AI factories.
Connects business model to infrastructure demand.
Unlock the full claims, asset map, scores, related transcripts, follow-up questions, and AI chat — shaped around your portfolio, watchlist, favorite speakers, and risks.
