U.S. Leadership in AI with Jensen Huang, Founder and CEO of NVIDIA, and Congressman Ro Khanna

What is necessary for the U.S. to maintain its leadership in AI?

How do we empower the workforce for the AI era?

NVIDIA Founder and CEO Jensen Huang and U.S. Congressman Ro Khanna (CA-17) met at Stanford for a conversation with H.R. McMaster on the future of technology and domestic policy.

It was moderated by former National Security Advisor General H.R. McMaster (with GSB Dean Sarah A. Soule also participating as host), focusing on maintaining U.S. AI leadership, empowering the workforce, regulating the technology responsibly, geopolitical competition (especially with China), reindustrialization, and sharing AI’s benefits broadly.

Maintaining U.S. Competitive Advantage

• Global Talent and Openness: Ro Khanna highlighted America’s edge in attracting worldwide talent to top universities and fostering diverse collaboration, which creates “magic and innovation.” He praised academic freedom, questioning authority, and government-university-private sector partnerships. Huang echoed the need for an open, welcoming environment for people from all backgrounds to uphold the American Dream.
• Reindustrialization and Manufacturing: Khanna called offshoring manufacturing a “colossal mistake” that harmed national security, social cohesion, and the Midwest (referencing his Pennsylvania roots). He advocated a “21st-century Marshall Plan” or affirmative jobs agenda to rebuild communities and share AI benefits. Huang noted AI is already aiding reindustrialization through data center construction and manufacturing needs, and agreed on the importance of maintaining an industrial base.

Workforce, Jobs, and Democratization of AI

Huang pushed back against fears that AI will eliminate jobs, using examples like radiology (demand for radiologists has increased despite automation) and his own role (AI automates typing/talking, yet he’s busier). He distinguished tasks (which can be automated) from job purposes and argued that people skilled in using AI are more likely to “take” jobs than AI itself.

He sees AI enabling shifts to higher-value work (e.g., carpenter to architect) and creating more overall employment. Both expressed optimism for students and the next generation: “You are exactly at the same place as everyone else. Nobody has a head start,” with every industry being reimagined and AI accessible to all.

Regulation, Risks, and National Security

• Regulation must balance innovation and safety without being premature, as technology evolves rapidly. McMaster compared it to asking the Wright brothers to write manuals for modern jets. Khanna wants U.S. standards for “excellent AI” (safe, privacy-respecting) to set global benchmarks. Huang warned against over-regulating or fearing AI out of society/industry.
• On geopolitics: Compete with China but not be “anti-China”; focus on cooperation alongside competition. Khanna and others emphasized building trust amid public skepticism tied to inequality.

California and Broader Optimism

Huang famously encouraged people to “Move to California. Don’t leave. It’s the highest taxes in the world, but it’s OK. The weather is great,” while being “perfectly fine” with taxes, prioritizing talent pools in Silicon Valley.

The tone was optimistic and forward-looking, with Huang especially encouraging students that the AI era resets opportunities equally and that the industry awaits their contributions to build the future. The panel stressed that AI’s success depends not just on tech but on policy, talent, reindustrialization, and equitable sharing of benefits.

The full ~1-hour video (judging by view counts and context) features engaging back-and-forth, with practical insights rather than deep technical dives. No full public transcript appears widely available yet, but the Stanford Daily and GSB recaps cover the essence comprehensively.

AI Infrastructure as a “Five-Layer Stack” (Huang’s Framework):

Jensen Huang described the AI industry as a multi-layered “cake” or stack: energy, chips/semiconductors, cloud infrastructure, AI factories (data centers), foundation models, and—most critically—the applications layer.

He argued that the U.S. must lead across all layers to maintain dominance, with applications being key to long-term economic impact and the “AI industrial revolution.” He stressed designing energy, compute, networking, and cooling together as integrated architectures for AI factories, noting surging demand for inference (token generation) and the ongoing massive buildout.

Jensen Huang (NVIDIA CEO) frequently describes AI as a “five-layer cake” or full-stack industrial system, emphasizing that it’s not just about models or software—it’s a complete infrastructure buildout akin to electricity or the internet, requiring massive real-world investment across every layer.

This framework highlights the interdependencies: each layer builds on the one below it, and every successful AI application ultimately pulls resources all the way down to the power plant. The U.S. (and companies like NVIDIA) must lead across the entire stack to maintain competitive advantage, with the top layer driving the most long-term economic value and job creation.

The Five Layers (Bottom to Top)

1. Energy (The Foundation/Binding Constraint): Intelligence generated in real time requires power generated in real time. AI data centers and training/inference have enormous electricity demands. Huang notes this as the ultimate limiter on scaling AI—China, for example, has roughly twice as much energy capacity as the U.S. in some contexts. Without abundant, reliable, and affordable energy (including new power plants, grids, and renewables/nuclear), the upper layers can’t scale. This layer drives massive infrastructure investment in utilities and power generation.
2. Chips (Processors/Compute Engines): Specialized hardware (like NVIDIA’s GPUs, along with CPUs, TPUs, networking chips, and high-bandwidth memory) that efficiently converts energy into massive parallel computation. AI workloads need high parallelism, fast interconnects, and specialized accelerators far beyond traditional CPUs. This is NVIDIA’s core strength, but the layer includes the entire semiconductor ecosystem (fabrication, design, advanced packaging).
3. Infrastructure (AI Factories / Cloud / Data Centers): The physical and systems layer: massive data centers (“AI factories”), networking, storage, cooling systems, software for orchestration, and cloud platforms that connect thousands or millions of chips into cohesive supercomputers. This includes building out facilities, power delivery, liquid cooling, and high-speed interconnects (e.g., NVIDIA’s NVLink or InfiniBand). Huang stresses designing energy, compute, networking, and cooling together as integrated architectures. This is where much of the current trillion-dollar buildout is happening.
4. Models (Foundation Models / AI Software): The trained large language models, multimodal models, and AI systems (e.g., GPTs, Grok, Llama, or domain-specific ones). These are the “intelligence” layer—trained on vast data using the lower layers, then used for inference (generating outputs). Models are important but not the end goal; they’re a reusable platform that powers applications.
5. Applications (The Most Important Layer for Economic Impact): The top layer where AI delivers real-world value: industry-specific tools, services, and transformations (e.g., AI in healthcare diagnostics, autonomous vehicles, drug discovery, creative tools, manufacturing optimization, or humanoid robots as “embodied applications”). Huang argues this is where the majority of jobs, productivity gains, and economic benefits will emerge—reimagining every industry. It’s the layer that “pulls” on all the ones below and determines the ultimate payoff of the entire stack.

Key Takeaways from Huang’s Framing

• Interdependence and Buildout Scale: AI isn’t lightweight software; it’s the largest industrial infrastructure project in history, involving energy, construction, manufacturing, and more. Success in one layer depends on all below it.
• U.S. Leadership: To dominate, America must excel in all layers (not just models or chips), including reindustrialization, talent attraction, and policy that supports the full stack. Applications will be key to broad prosperity and reindustrialization.
• Optimism on Jobs: Every layer creates new work—from power plants and chip fabs to data center construction and AI-enhanced roles in traditional industries.
• Dynamic Nature: The stack is being reinvented in real time; inference demand is surging, and new applications (including physical ones like robotics) will drive further growth.

This model has been a recurring theme in Huang’s 2025–2026 talks (including Davos, CSIS, and the Stanford event with Ro Khanna). It shifts the conversation from “AI chatbots” to a profound industrial transformation. For the most direct explanation, see NVIDIA’s blog post “AI Is a 5-Layer Cake.”