An Interview with Nvidia CEO Jensen Huang About Accelerated Computing
Huang explains why Nvidia is expanding beyond chips into full-stack computing and AI agents.
In a post-GTC 2026 interview, Nvidia CEO Jensen Huang framed the company's keynote as a "history lesson" to reinforce its core identity: a full-stack accelerated computing platform, not just a chip supplier. He emphasized that as AI evolves to use tools agentically—like Excel, Photoshop, and CAD software—those tools must be "super-accelerated" to keep pace. This requires Nvidia to expand CUDA and its software libraries into new verticals, exemplified by a newly announced partnership with IBM for data processing.
Huang addressed Nvidia's strategic moves, including its growing CPU business and the acquisition of Groq, framing them not as a change in direction but as a natural extension of its accelerated computing mission. He pushed back against the notion that Nvidia is being "squeezed into the chips box," stating his focus is on "what we need to be" to solve full-stack problems. The discussion also covered the scarcity in the AI stack, challenges in China, and Huang's frustration with AI "doomers" influencing policy.
- Huang states AI agents will require "super-accelerated" versions of human tools like Excel and Photoshop, driving Nvidia into new software verticals.
- Nvidia's strategy includes expanding partnerships (like with IBM) and acquisitions (Groq) to build a full-stack accelerated computing platform beyond just chips.
- The CEO rejects the idea that Nvidia is just a chip company, emphasizing a 20-year focus on solving full-stack problems through CUDA and software libraries.
Why It Matters
Signals Nvidia's move to dominate the infrastructure layer for the coming wave of AI agents, beyond just selling GPUs.