“We Cannot Drink Data”
AI may feel digital, but its growth depends on very physical resources. A closer look at the power, water and infrastructure constraints sitting underneath the AI boom.
Taheera Lovell
8/3/20261 min read
There is something wonderfully misleading about the word cloud...it sounds weightless.
In reality, AI depends on some very physical things:
Power.
Water.
Land.
Data centres.
Chips.
Transmission networks.
Cooling systems.
And increasingly, the AI boom is colliding with the limits of that infrastructure.
That changes the conversation.
The race for AI leadership may not only be about who has the best models.
It may also be about who can provide enough electricity, grid connections, compute capacity and water to run them.
Which is where my resilience brain starts twitching.
Because digital ambition eventually runs into physical constraints.
A country may have brilliant researchers and ambitious AI policy.
But if it cannot build enough generation, transmission or data-centre capacity, physics gets a vote.
Water adds another layer.
Data centres exist somewhere.
They use local infrastructure.
They may compete with housing, agriculture, industry and communities for resources.
So “how much water does AI use?” may be less useful than: Where is the computing happening, what resources exist there, and who else needs them?
This is why the AI boom is also an infrastructure story.
And possibly an energy story.
A water story.
A planning story.
A sovereignty story.
Some of the most strategically important AI investments may not have “AI” written on them at all.
They may be grids, substations, cooling systems, generation, storage and water infrastructure.
The cloud was never really in the sky.
We are just starting to notice what it is standing on.