AI Is a 5-Layer Cake — And Every Business Needs to Understand It

AI Is a 5-Layer Cake — And Every Business Needs to Understand It

Most people think about AI the way they think about software — as an app you download, a feature you enable, or a tool you subscribe to. Jensen Huang, NVIDIA's founder and CEO, wants to fundamentally change that mental model. In a recent post on the NVIDIA blog, he lays out a framework that every founder, investor, and policymaker should internalize: AI is infrastructure, and it runs on five tightly coupled layers.

The Shift That Changes Everything

For most of computing history, software was prerecorded. Humans wrote algorithms. Computers executed them. Data had to be carefully structured and stored in tables, retrieved through precise queries like SQL.

AI breaks that entirely.

For the first time, we have a computer that can understand unstructured information — images, text, sound and meaning. It can reason about context and intent. Most importantly, it generates intelligence in real time. NVIDIA Newsroom

That last part is the key insight. Intelligence produced in real time means the entire computing stack beneath it had to be reinvented from scratch.

The Five Layers

Jensen describes AI as a five-layer cake, where each layer depends on everything beneath it:

1. Energy At the foundation is power. Every token an AI generates is the result of electrons moving and heat being managed. There is no abstraction layer beneath this. Energy is the first principle of AI infrastructure and the binding constraint on how much intelligence the system can produce.

2. Chips Above energy are processors designed to transform that power into computation at massive scale. Progress at the chip layer determines how fast AI can scale and how affordable intelligence becomes — which is exactly why NVIDIA's Rubin, Blackwell, and future architectures matter so much.

3. Infrastructure This is where the physical world meets AI: land, cooling, networking, power delivery, and the systems that orchestrate tens of thousands of processors into one machine. These systems are AI factories. They are not designed to store information. They are designed to manufacture intelligence.

4. Models Above infrastructure are the models — systems trained to understand language, biology, chemistry, physics, finance, and the physical world. Language models are just one category. Some of the most transformative work is happening in protein AI, chemical simulation, robotics, and autonomous systems.

5. Applications At the top is where economic value is created: drug discovery platforms, industrial robotics, legal copilots, self-driving vehicles. A self-driving car is an AI application embodied in a machine. A humanoid robot is an AI application embodied in a body. Same stack. Different outcomes.

Why Open Models Accelerate Everything

One underappreciated point in Jensen's framework is the role of open source models. Most of the world's AI models are freely available. When open models reach frontier performance, they don't just change software — they activate demand across every layer of the stack beneath them.

DeepSeek-R1 was a powerful example of this. By making a strong reasoning model widely available, it accelerated adoption at the application layer and increased demand for training, infrastructure, chips and energy beneath it.

This is a virtuous cycle. Better open models → more applications → more infrastructure demand → more investment in chips and energy → even better models.

The Scale of What's Coming

Here's the number that should stop you mid-scroll: we are a few hundred billion dollars into the AI buildout. Trillions of dollars of infrastructure still need to be built.

This is why Jensen calls it the largest infrastructure buildout in human history — bigger than railroads, bigger than the internet, bigger than the electrical grid.

And critically, it's not just a job for PhDs. AI factories need electricians, plumbers, pipefitters, steelworkers, network technicians, installers and operators. These are skilled, well-paid jobs, and they are in short supply.

What This Means for Africa

For builders and investors on the African continent, Jensen's framework is a roadmap, not just a story about Silicon Valley.

Nigeria, Kenya, Rwanda and South Africa are all making moves on AI infrastructure. The question isn't whether to participate — it's which layer to enter. Most African startups are rightly focused on the application layer, where the economic value is created and the capital requirements are lowest. But the bigger opportunity over the next decade will be in model customisation and infrastructure deployment — AI factories built for African languages, African data and African problems.

SmartSchoolAfrica is one example of this: a platform sitting at the model and application layers simultaneously, using fine-tuned AI to generate job-ready curriculum from raw documents. The five-layer cake framework explains exactly why that kind of African-first application layer investment matters — every successful application pulls demand all the way down to the power plant.

The Bottom Line

AI is becoming the foundational infrastructure of the modern world. And the choices we make now — how fast we build, how broadly we participate and how responsibly we deploy it — will shape what this era becomes.

The five-layer cake isn't just a useful mental model. It's a checklist. If you're building in AI, ask yourself: which layer am I in? What does the layer below me look like? Who controls it? What happens if it breaks?

Every company will use AI. Every nation will build it. The question is whether you'll be a consumer of that infrastructure — or a builder of it.

Source: "AI Is a 5-Layer Cake" — Jensen Huang, NVIDIA Blog, March 10, 2026