Where AI Is Actually Built: Chips, Power and Land

The cloud is the most successful metaphor in modern business and one of the most misleading. Artificial intelligence is not built in a cloud. It is built on the ground, in windowless buildings the size of cathedrals, wherever three scarce things happen to converge: advanced chips, cheap and abundant electricity, and a way to carry away the heat. Understand that convergence and you understand why the geography of intelligence is far narrower, and far more physical, than the word cloud implies.

The Physical Trinity

Every capable AI system rests on a trinity. First, the chips, a class of processors so difficult to make that the world’s supply passes through a handful of facilities and depends on machines produced by essentially one company. Second, power, because training and running these systems consumes electricity on the scale of small cities, continuously, without the daily rhythm that lets ordinary demand rest at night. Third, cooling, because everything that consumes that much power converts it into heat, and heat must go somewhere, usually into air, increasingly into water.

None of these three is evenly distributed, and that unevenness is the whole story. Chips have a bottleneck of manufacturing. Power has a bottleneck of generation and grid. Cooling has a bottleneck of climate and water rights. Intelligence gets built only where all three bottlenecks happen to open at the same place, and there are not many such places on Earth.

Why the Chip Is a Chokepoint, Not a Commodity

We are used to thinking of components as commodities, interchangeable and cheap. The leading AI processor is the opposite: a chokepoint. It is fabricated using a lithography process so precise that only one supplier in the world makes the machines that make it, and only a few fabrication plants can run those machines at scale. This concentration means that the physical capacity to build advanced AI is gated not by ideas or money but by a production line that cannot be hurried and cannot be easily copied.

This is why the chip has become an instrument of statecraft. When something sits at a chokepoint, controlling its flow becomes a form of power, and nations treat access to it the way earlier eras treated access to sea lanes or oil. The important insight is not political, though; it is structural. Whenever a technology depends on a single narrow passage, that passage becomes the place where leverage concentrates, and everything downstream reorganizes around it.

Power Is the New Ceiling

For a while the binding constraint was chips. It is quietly becoming power. You can, with enough patience and capital, acquire processors; you cannot conjure a gigawatt of reliable electricity where the grid does not reach. This is why the builders of AI have started behaving like the builders of aluminum smelters a century ago: they go to the electricity rather than making the electricity come to them.

That behavior explains a migration that surprises people. The map of where AI gets built increasingly overlaps with the map of stranded or abundant power: places with surplus hydroelectricity, with cheap gas, with land beside a reactor, with wind that has nowhere else to go. The intelligence follows the electrons. And because power infrastructure takes years to build and decades to pay off, the sites chosen now will shape the industrial geography of the next generation.

Cooling, Water, and the Return of Climate

The third constraint is the one most people forget, and it is the one that ties intelligence back to the oldest kind of geography: climate and water. A dense cluster of processors is a furnace. Traditionally the heat went into the air, but at scale that becomes inefficient, so operators increasingly move heat into water, which means they need water, and clean, reliable water is not everywhere. Cool climates help. Rivers and aquifers help. Suddenly a technology that felt weightless is negotiating for the same resource that decided where ancient cities rose.

Here is the one comparison worth keeping. A data campus is, in its logic, closer to a nineteenth-century textile mill than to a software company. The mill went where the river fell, because falling water was power and moving water was cooling and transport. It could not choose its location freely; the geography chose it. The AI campus is the same creature in modern dress: it goes where power falls and water flows, and it, too, is chosen by the land more than it chooses.

The Land Underneath

All of this lands, finally, on real ground. The convergence of chip supply, power access, and cooling water expresses itself as a hunger for very specific sites: parcels beside substations, plots near cool water, jurisdictions willing to permit heavy load and heavy build quickly. These sites are scarce because the overlap of conditions is scarce, and scarce things beside unavoidable infrastructure have always been where quiet value forms.

The lesson for anyone reading the future is to stop imagining AI as ethereal and start seeing it as heavy industry with a soft interface. The interface is weightless. The industry beneath it is as physical as steel, and like all heavy industry it stamps itself onto particular places and stays there.

FAQ

Why can’t AI just be built anywhere, given that it runs on the internet? Because the internet is only the interface. The computation itself needs advanced chips, enormous continuous power, and large-scale cooling, and those three requirements overlap in only a few locations. The network is everywhere; the factory of intelligence is not.

Is the main constraint chips or electricity? It is shifting from chips to electricity. Chips are a manufacturing bottleneck that capital and time can ease. Power is a physical bottleneck: you cannot place a gigawatt of reliable generation where the grid and the fuel are not, which is why builders now migrate toward abundant power.

Why does cooling matter so much? Because everything that consumes that much power becomes heat, and at scale the heat is moved into water. That reintroduces an ancient constraint, access to reliable water and a cool climate, and quietly ties a futuristic industry back to the oldest logic of geography.

At Kev Living we read the future by following the electrons and the water to the ground they favor, because that is where the next map of value is being poured.