Layer 1 · Physical & Energy
The power, cooling, buildings and chip factories that everything else physically runs on.
In simple words: The power stations, cooling and chip factories that every AI needs before it can think at all.
Like: The land, roads and power lines of a city: invisible until they run out.
Where it breaks: Roads can be widened in months. Chip factories and grid connections take years and enormous investment, and advanced chip production is concentrated in very few places, so this layer can bottleneck the whole stack at once.
Key players
What it means for your business: You will rarely buy from this layer directly, but power and chip-factory limits decide how much AI capacity exists, which shapes cloud prices and how quickly providers can grow.
- TSMC: World-leading contract chip factory; builds the chips NVIDIA and others design.
- ASML: The only maker of the EUV machines needed to print leading-edge chips.
- Samsung Foundry: Korean chip factory that also builds chips for outside designers.
- Constellation Energy: Largest US nuclear operator; sells long-term power to data centres.
- Equinix: Rents data-centre space and connections in 36 countries.
- Schneider Electric: French maker of power and liquid-cooling systems for AI data centres.
- Vertiv: Supplies power and cooling equipment for dense AI racks.
What it gives the layer above
Electricity, cooling and floor space, plus the fabricated chips themselves.
What it needs from the layer below
Nothing in the AI stack: it rests on energy markets, land, water and industrial supply chains.
Concepts
- Grid connection & energy supply
A big AI data centre needs a large, steady supply of electricity, and getting connected to the grid can take years, so power, not chips, often decides where and when one gets built. - Power & cooling for AI data centres
AI data centres are limited by how much electricity they can get and how much heat they can remove. - Liquid cooling
AI racks pack so much power into a small space that fans and cold air can no longer carry the heat away, so water-based cooling is piped right up to the chips. - Fabs & the chip supply chain
Advanced AI chips depend on a handful of highly specialised factories, tools and materials. - Energy per token
Each answer a model gives costs a little electricity, and how much depends on what you count: the chips, the rest of the server, idle machines and the building's cooling.
Walkthrough
- An answer starts as electricity. Someone asks a chatbot a question. Before any maths happens, the data centre needed a grid connection big enough for a town, which took years to secure.
- Power in, heat out. The electricity flows through substations to racks of AI chips, and nearly all of it turns into heat while the chips work out the answer.
- Liquid carries the heat away. Those racks are so dense that air cannot keep up, so liquid flows through plates on each chip and carries the heat outside.
- Chips from a handful of factories. The chips doing the work came from a few specialised fabs and packaging lines, which is why supply can run short.
- The bill for one answer. Add up the chips, the server, idle machines and cooling, and you get the energy per token. Making that smaller stretches every megawatt the grid can supply.
Examples
- Data-centre campuses
- Grid connection queues
- Direct-to-chip liquid cooling
- Semiconductor fabs
- Energy per token and PUE