| Episode | Status |
|---|---|
| Episode | Status |
|---|---|
Even if ChatGPT never existed, the tech giant NVIDIA would still be winning. The end of Moore’s Law—says NVIDIA President, Founder, and CEO Jensen Huang—makes the shift to accelerated computing inevit...
NVIDIA CEO Jensen Huang discusses 2025's AI breakthroughs including reasoning improvements, profitable inference tokens, and grounding advances. He refutes AI bubble narratives by explaining the fundamental shift from general-purpose to accelerated computing, emphasizes the critical importance of open source for startups and industry innovation, and addresses job displacement fears by distinguishing between task automation and job purpose. Huang predicts major 2026 breakthroughs in digital biology, autonomous vehicles with reasoning capabilities, and humanoid robotics while advocating for pragmatic US-China relations and energy policy to support AI infrastructure growth.
Huang reflects on 2025's key achievements including major improvements in AI reasoning, grounding, and search integration. Most significantly, inference tokens—especially reasoning tokens—became highly profitable with companies like OpenEvidence achieving 90% gross margins, validating the business model of AI applications.
Huang explains how AI's unique requirement for continuous token generation is creating three new types of manufacturing facilities: chip plants, supercomputer plants, and AI factories. This infrastructure buildout is generating enormous demand for skilled labor including electricians, construction workers, and network engineers, with wages doubling in some cases.
Huang introduces a critical framework distinguishing between job 'tasks' (what you do) and job 'purpose' (why you do it). Using radiologists as an example, he shows how AI automating tasks actually increased radiologist employment because it enabled them to better fulfill their purpose of diagnosing disease, making hospitals more productive and profitable.
Huang argues robotics will primarily address severe labor shortages in manufacturing, trucking, and other industries rather than displacing workers. He emphasizes that aging populations globally are creating workforce gaps that automation must fill, and that robots will create massive new repair and maintenance industries.
Huang provides a comprehensive defense of open source AI, explaining the five-layer technology stack (energy, chips, infrastructure, AI models, applications) and why open source is essential for startups, research, education, and established industries. He warns policymakers against damaging the open source innovation flywheel.
Huang strongly criticizes 'doomer' narratives about AI, calling them 'extremely hurtful' and suggesting they may be motivated by regulatory capture to suffocate competition. He argues that rapid AI advancement actually solved safety problems like hallucination and grounding, and that the first principle of safety is that products work as advertised.
Huang explains the dramatic cost reductions in AI, noting that the first ChatGPT now costs essentially nothing to build. He projects billion-times cost reduction over ten years through compounding improvements in hardware (5-10x/year), algorithms, and model architecture, making the 'AI is too expensive' narrative misleading.
Huang credits DeepSeek as 'probably the single most important paper that most Silicon Valley researchers read' in recent years, arguing it was the greatest contribution to American AI in 2025. This demonstrates how open research from other nations benefits US innovation and why isolationist thinking is counterproductive.
Huang predicts several industries will have their 'ChatGPT moment' in 2026, particularly digital biology with breakthroughs in protein synthesis and multi-protein understanding. He also forecasts major advances in autonomous vehicles and humanoid robots through reasoning systems that handle out-of-distribution scenarios.
Huang explains why robotics will advance faster than self-driving cars did, having learned from that ten-year journey. He emphasizes that AI will be multi-embodiment (like humans operating different tools) and that 'everything that moves will be robotic,' creating diverse opportunities beyond just humanoid robots.
Huang credits the Trump administration's energy policy shift as preventing the US from 'handing this industrial revolution to somebody else.' He argues that without energy growth, there can be no industrial growth, and that AI infrastructure demand is ironically driving the biggest climate innovation boom in history.
Huang expresses optimism about US-China relations under the Trump administration, advocating for a nuanced strategy that recognizes both countries are adversaries and partners. He supports export controls grounded in national security while acknowledging deep economic coupling and the importance of American technology leadership globally.
Huang systematically dismantles AI bubble concerns by explaining that even without chatbots, NVIDIA would be a multi-hundred-billion dollar company due to the fundamental shift from general-purpose to accelerated computing. He shows how global R&D spending ($2 trillion annually) is shifting from wet labs to AI supercomputers.
Huang concludes by emphasizing the importance of understanding AI as a multi-layer technology stack spanning energy to applications, and recognizing its diversity across domains beyond chatbots. This framework helps make narratives more pragmatic and balanced, showing why America should aim to win across all layers and domains, not just one company or application.
NVIDIA’s Jensen Huang on Reasoning Models, Robotics, and Refuting the “AI Bubble” Narrative
Ask me anything about this podcast episode...
Try asking: