What's next
The roadmap he preaches — and the questions he can't answer.
The roadmap he preaches
Huang’s public vision for the next decade has three pillars:
1. Physical AI and robotics
After language models (“digital AI”), Huang argues the next wave is physical AI — robots and autonomous systems that understand the physical world. NVIDIA’s Omniverse (simulation/digital twins) and Isaac (robotics platform) are the bets: train robots in simulation, deploy them in reality. The GTC demos increasingly feature humanoid robots and self-driving systems.
2. Sovereign AI
Every nation, Huang argues, should build its own AI infrastructure — data centers, models, talent — rather than importing intelligence. He has carried this pitch to governments across Asia, the Gulf, and Europe. It’s part industrial policy, part sales strategy, and entirely consistent with his worldview: computing infrastructure as national infrastructure, like electricity or railways.
3. The relentless cadence
Blackwell now, Rubin next (announced for 2026, named for astronomer Vera Rubin), and another architecture after that. The strategy is the treadmill: stay one full generation ahead, make each generation obviously worth the upgrade, and let CUDA’s software moat do the rest.
The questions he can’t answer
Succession. Huang has been CEO since 1993 — one of the longest founder-CEO runs in business history. He is in his sixties. NVIDIA’s entire management system (sixty direct reports, no planning cycle, decisions by founder reasoning) is built around him. What the company looks like without him is the most important unanswered question in tech, and there is no public answer.
The cycle. Every computing boom has a digestion period. If hyperscaler capex pauses, NVIDIA’s growth story changes overnight. Huang’s answer — “thirty days from going out of business” — is culture, not a hedge.
Geopolitics. Export controls, Taiwan risk, sovereign-AI deals that are also diplomatic entanglements: NVIDIA sits at the intersection of U.S.–China tech competition. No keynote controls that.
Energy. “We are a power-limited industry,” Huang said at GTC 2025. The AI buildout’s electricity demand is an engineering constraint and a political one.
The honest summary
The roadmap is clear and the risks are real — and both are visible in the same keynotes. Huang’s bet is that robotics and sovereign AI extend the buildout for another decade; the market’s bet is that he’s right; history’s bet is that no dominance lasts forever. This page will look either prescient or quaint in five years. That’s the nature of “what’s next.”
Source notes
Roadmap items draw on GTC 2024–2025 keynote coverage and Huang’s interviews on robotics and sovereign AI. Succession, cycle, geopolitical, and energy risks are characterized from financial and policy press coverage; no succession plan has been publicly disclosed.