Jensen Huang of NVIDIA is quite the worldwide AI leader, up there with Altman in determining where AI ends up.
Was sort of interviewed by BlackRock’s Larry Fink at DAVOS. Fink contributed a few short sentences…one of which was to say he thinks Pension Funds should invest in the Infrastructure layer…so that the common man can get in on it. (Also partly where NVIDIA sells its chips). Mr. Fink tried a few times to broaden the AI development conversation to include goals of what’s best for society and not just what’s best for share prices. But that discussion was minimal at best.
There’s what Huang’s saying…and there’s why he’s saying it…
….and then there’s what he isn’t saying.
Nvidia CEO Jensen Huang speaks at the World Economic Forumyoutube.com
Huang also made a point saying it two different times that VC spending in 2025 was at record levels and heavily tilted toward layer 5…where the big productive output shows up. Such as Healthcare tools.
That 5th layer is seemingly a more active layer for NVIDIA than layer 4 where he doesn’t quite have the LLMs to compete.
Huang does surprisingly have some advanced self driving car tools which might live in the top layer too.
Huang’s “five-layer cake”
Bottom layer: Energy
The electricity required to run AI in real time (training and especially inference at scale).
Layer 2: Computer chips and computing infrastructure
The hardware layer Huang “lives in”: GPUs/CPUs, networking, racks/servers, and the broader accelerated-computing infrastructure.
Layer 3: Cloud infrastructure (cloud services / data centers)
The cloud/data-center layer that operates and delivers the compute as services (the “AI factories” side of the stack).
Layer 4: AI models
The foundation models and related systems most people think of as “AI” (LLMs, multimodal, etc.).
Top layer: Applications
The app layer (finance, healthcare, manufacturing, etc.) where the economic benefit ultimately shows up.
John