Speaking at the World Economic Forum in Davos, NVIDIA founder and CEO Jensen Huang positioned artificial intelligence not as a single technology, but as the foundation of what he described as the largest infrastructure buildout in history.
In a discussion with BlackRock CEO Larry Fink, Huang repeatedly returned to the idea that AI should be understood as a foundational layer of modern economies, rather than a discrete software wave. “AI is infrastructure,” Huang said, arguing that every country should treat it in the same way it treats electricity or roads. In his view, national competitiveness will increasingly depend on whether countries build and operate their own AI capabilities, rather than relying entirely on external platforms.
That framing places AI alongside energy, transport, and communications as critical infrastructure. Huang expanded on this by describing AI as a multi-layer system rather than a single breakthrough. He characterised it as a “five-layer stack,” spanning energy and power generation, chips and computing infrastructure, cloud data centres, AI models, and finally the application layer. “AI is not one thing,” Huang said. “It’s a five-layer stack, from energy and computing infrastructure all the way up to the application layer, where the economic benefit ultimately happens.”
Each of those layers, Huang argued, must be built, operated, and maintained, which is already driving demand across a wide range of industries. From chip manufacturing and data centre construction to network operations and application development, the AI buildout is reshaping labour needs across both skilled trades and advanced engineering roles.
While public attention often focuses on AI models and applications, Huang emphasised the physical foundations required to support them. Energy generation, semiconductor manufacturing, and large-scale computing infrastructure are not side concerns, but prerequisites. In that sense, the AI boom is as much about physical systems as it is about software.
Huang also pushed back against the idea that AI adoption will inevitably reduce employment. Instead, he described a shift from task-based work to purpose-driven roles. In healthcare, for example, AI is increasingly used to handle administrative and repetitive tasks, allowing professionals to spend more time on diagnosis, patient interaction, and decision-making. In radiology and nursing, he argued, productivity gains have coincided with continued or increased demand for skilled workers.
That perspective aligns with Huang’s broader argument that AI is lowering barriers to entry rather than raising them. He noted how quickly AI tools have spread globally, reaching hundreds of millions of users in just a few years. As a result, AI literacy is becoming a baseline skill, not a specialist one, encompassing not just how to use models, but how to direct, evaluate, and govern them responsibly.
Turning to manufacturing and robotics, Huang highlighted what he called a once-in-a-generation opportunity for regions with strong industrial bases. Rather than treating AI as something that is simply written or coded, he urged companies and countries to focus on teaching AI through real-world processes, data, and environments. This fusion of industrial expertise and artificial intelligence, he suggested, is what will unlock progress in physical AI and robotics.
The message from Davos was not that AI adoption is slowing, but that it is becoming more infrastructure-heavy, more capital-intensive, and more tightly coupled to physical systems. As Fink observed in closing, the key question may no longer be whether AI represents a speculative bubble, but whether governments, industry, and investors are moving quickly enough to build the layers required to support it.
That question echoes themes raised elsewhere at Davos, including in our recent analysis of Elon Musk’s remarks on robots, autonomy, and scale. In that piece, [More Robots than People, What Elon Musk’s Message Means for Vision], we explored how ambitious claims about AI-driven abundance ultimately depend on perception systems, physical reliability, and deployment realities.
Taken together, the conversations point to a shared conclusion. Whether the focus is autonomy, robotics, or industrial AI, the next phase of progress will be shaped less by abstract models and more by the physical and infrastructural layers that make those models usable in the real world.
