Jensen Huang at WEF: The Largest Infrastructure Buildout in Human History

AI is the largest infrastructure buildout in human history. We are a few hundred billion dollars into it. Trillions more are needed. And this is not a bubble; it is a platform shift on the scale of the PC, the internet, and mobile computing. That is the central thesis Jensen Huang brought to his first appearance at the World Economic Forum in Davos, in a conversation hosted by BlackRock CEO Larry Fink.

The Conversation

Larry Fink opens with a striking comparison: since both NVIDIA and BlackRock went public in 1999, NVIDIA has delivered a 37% compounded annual return versus BlackRock’s 21%. Jensen’s one regret? Selling NVIDIA stock at a $300 million valuation to buy his parents a Mercedes S-Class. “It is the most expensive car in the world,” he jokes. They still have it.

The conversation then moves through AI’s industrial architecture, its impact on jobs, its potential for developing economies, and Europe’s specific opportunity, all in about 30 minutes.

The Five-Layer Cake

Jensen frames AI not as a single technology but as a complete computing stack with five layers:

1. Energy — AI processes intelligence in real time and needs energy to do so. This is the foundation.
2. Chips and computing infrastructure — “The layer I live in.” TSMC is building 20 new chip plants. Micron is investing $200 billion in the United States. SK Hynix and Samsung are scaling aggressively.
3. Cloud infrastructure — The cloud services layer that hosts and distributes compute.
4. AI models — “This is where most people think AI is.” But without the layers beneath, models cannot exist.
5. Applications — The layer where economic value is ultimately realized: healthcare, finance, manufacturing, robotics.

The key insight is that most public attention is fixed on layer four (models), but all five layers must be built simultaneously. This is why the infrastructure investment is so massive and why it is not speculative.

“We’re now a few hundred billion dollars into it. There are trillions of dollars of infrastructure that needs to be built out.”

Three Breakthroughs of 2025

Jensen identifies three developments that made 2025 a breakthrough year for AI at the model layer:

Agentic AI. Language models evolved from curious-but-hallucinating chatbots into systems that can reason step by step, plan, and execute tasks. The shift from “language models” to “AI systems” is the defining transition.

Open models. DeepSeek’s release was the pivotal moment. Jensen reframes the panic around it: “DeepSeek was a huge event for most of the industries, most of the companies around the world because it’s the world’s first open reasoning model.” Open models enabled companies, researchers, educators, and startups to build domain-specific AI without starting from scratch.

Physical AI. AI that understands not just language but the physical world: proteins, chemicals, fluid dynamics, particle physics. The Eli Lilly partnership is the flagship example. AI can now “interact and talk to proteins like we talk to ChatGPT,” and drug discovery is seeing real acceleration.

Purpose vs. Task: A Framework for Job Impact

The most concrete part of the conversation is Jensen’s argument that AI creates jobs rather than eliminates them. He builds this on a simple framework: distinguish between the purpose of a job and the task of a job.

Radiology. Ten years ago, radiology was the profession everyone expected AI to wipe out, because computer vision was the first AI capability to reach superhuman performance. Ten years later, AI has “completely permeated every aspect of radiology.” The impact is 100% real. And the number of radiologists has gone up. Why? Because the purpose of a radiologist is to diagnose disease, not to study scans. With scans processed instantly, radiologists spend more time with patients. Hospitals can see more patients. Revenue goes up. They hire more radiologists.

Nursing. The U.S. is 5 million nurses short. Nurses spend half their time charting and documenting. AI (Jensen mentions Abridge as a partner) automates the charting. Nurses spend more time with patients. Hospitals can see more patients. They hire more nurses.

“If you just put a camera on the two of us and just watched us, you would probably think the two of us are typists. Because I spend all of my time typing. And so if AI could automate word prediction and help us type, then we would be out of jobs. But obviously that’s not our purpose.”

Jensen argues this framework generalizes: for any job, ask what is its purpose (caring for patients, diagnosing disease, making investment decisions) versus what are its tasks (studying scans, charting, data retrieval). AI automates the tasks, which amplifies the purpose, which increases demand for the humans who fulfill that purpose.

The infrastructure buildout itself is also a massive job creator. Plumbers, electricians, construction workers, steel workers, network technicians. U.S. salaries in these trades have nearly doubled, reaching six figures. “You don’t need a PhD in computer science to make a great living.”

Every Country Needs AI Infrastructure

On the developing world, Jensen makes a pointed argument: AI is infrastructure, like electricity and roads. Every country should build it, not just import it.

The logic is straightforward. Open models have made AI accessible. Local expertise can fine-tune models for local languages and cultures. “Your fundamental natural resource is your language and culture,” he tells the Davos audience. Countries should develop their own AI, refine it, and treat national intelligence as part of their ecosystem.

He also points out that AI is “the easiest software to use in history,” which is why adoption is approaching a billion users in just two to three years. The barrier to entry has never been lower.

“If you don’t know how to use an AI, just go up to the AI and say, I don’t know how to use an AI. How do I use an AI? And it would explain it to you.”

Jensen specifically praises Claude: “Anthropic has made a huge leap in developing Claude. We use it all over our company. The coding capability, its reasoning capability… anybody who’s a software company really ought to get involved and use it.”

Europe’s Robotics Moment

For Europe specifically, Jensen sees a once-in-a-generation opportunity rooted in its industrial manufacturing base. His argument:

• The U.S. led the era of software. Europe largely missed it.
• But AI changes the rules: “AI is software that doesn’t need to write software. You don’t write AI, you teach AI.”
• Europe’s deep industrial base, its manufacturing capability, and its strong tradecraft workforce position it perfectly for physical AI and robotics.
• Europe’s deep sciences remain world-class and can now be accelerated by AI.
• The critical prerequisite: Europe must invest seriously in energy supply to power the infrastructure layer.

The tradecraft workforce point is notable. Jensen observes that the U.S. “lost that in the last 20, 30 years” while it remains strong in Europe. This is an asset, not a liability, in the AI infrastructure era.

Not a Bubble

Larry Fink raises the bubble question directly. Jensen’s response is empirical: NVIDIA has millions of GPUs in the cloud. If you try to rent one, the spot price is going up, not just for the latest generation but for GPUs two generations old. Demand is outstripping supply.

The investment data supports this. 2025 was one of the largest VC funding years ever, exceeding $100 billion globally. Most of that funding went to “AI native companies” in healthcare, robotics, manufacturing, and financial services, companies building the application layer above.

Eli Lilly is the corporate example: R&D budget that used to go entirely to wet labs is increasingly shifting to AI supercomputers.

Larry Fink adds that AI infrastructure should be a major investment direction for global pension funds. “We need to make sure that the average pensioner, the average saver is part of that growth.”

Jensen closes with a direct call to action:

“This is the single largest infrastructure buildout in human history. Get involved.”

Some Thoughts

This is a deliberately accessible conversation. Jensen is not revealing new technical breakthroughs; he is building an investment thesis for a Davos audience of political leaders, pension fund managers, and industrial executives.

• The five-layer stack is simple but genuinely useful. It explains why AI investment must be massive (five layers, not one) and why returns at the application layer depend on all the layers beneath being built out.
• The purpose-vs-task framework for job impact is the strongest section. It is testable (radiology numbers are real), it generalizes, and it gives policymakers a concrete tool beyond “AI replaces jobs.”
• Jensen’s praise of Claude is notably specific. Publicly naming a product in Anthropic’s ecosystem (NVIDIA works with all AI companies) suggests Claude’s position in enterprise coding and reasoning has been validated at the top of the industry value chain.
• The most striking data point: tradecraft salaries in the U.S. have “nearly doubled” to six figures for people building chip factories, computer factories, and AI factories. The AI economy is already creating tangible wealth for workers without computer science degrees.
• The “most expensive car in the world” joke captures something real about opportunity cost. Selling NVIDIA at a $300 million valuation to buy a Mercedes. His parents still drive it.