The Cloud
Drinks First on AI infrastructure, water, and who absorbs the cost of computation

The cloud is a bad metaphor. The better question is: through whose water is AI being cooled? That is where this essay begins. Not because metaphor does not matter, but because metaphor is one of the places policy hides. "The cloud" is a sky word for a ground problem. It makes computation feel atmospheric, weightless, inevitable. It keeps your eyes lifted while the real story stays below the line of sight: pumps, chillers, substations, permits, withdrawals, zoning hearings, tax abatements, utility load forecasts, and the neighborhoods drafted into consequence without authorship.

This is not written from a neutral infrastructure lens. It is written from a Black ecological and worldmaking lens that treats metaphor, governance, and material survival as inseparable. That means less interest in whether a technology can be described as transformative than in what it asks land, water, and public systems to absorb so that transformation can be claimed.

The language of "the cloud" is one way infrastructure teaches organized forgetting.

AI buildout is moving faster than most public processes are designed to metabolize. Hype travels at platform speed. Permits move quietly. Utility planning happens in technical language. Water stress is discussed as a regional condition, while the burdens of cooling infrastructure are negotiated locally, often in places already carrying layered exposure. By the time a project becomes publicly legible, many of its terms have already been shaped.

This essay is also for cultural workers, because what often gets called technopoetics at the level of form is always shadowed by infrastructure at the level of material support: grids, cooling systems, water withdrawals, platform economics, and public systems asked to underwrite private tools. If we do not read those layers together, we mistake style for autonomy.

Because the cloud does not float. It sits somewhere. On land. Near a watershed. Inside a grid. In a county with a planning board. In a city with a mayor. In a utility territory. Near people who usually do not get to define the timeline of its arrival.

And because it sits, it pulls. Water. Power. Public attention. Administrative flexibility. Land use. Capital. It pulls all of that into private acceleration and asks to be called the future.

Recent reporting has made the scale harder to ignore: large data centers can require immense amounts of water for cooling, and many new facilities are being developed in water-stressed regions. Those are not just technical figures. They are political measurements. They tell us the argument is no longer only about software capability. It is about who bears the material costs of computation.

The sharper question is how benefit and burden are distributed through different infrastructures and different timelines. Hyperscale firms secure compute capacity and competitive advantage. Developers and financiers capture land and infrastructure value. Municipal officials secure growth narratives and promised revenue. Utilities gain large-load demand and leverage for system expansion.

Residents, meanwhile, absorb slower and less marketable costs: water strain, heat vulnerability, air quality risk, noise, public-service pressure, and the cumulative stress of systems organized around someone else's urgency. That is the geometry.

If this still sounds abstract, look at South Memphis. The conflict around xAI's supercomputer facility near predominantly Black communities, including Boxtown, makes the pattern visible. One side is permitted to speak in the language of jobs, investment, competitiveness, and momentum. The other is forced to prove, again, that air, water, health, and neighborhood survival count as "real" infrastructure.

And because Boxtown's struggle is shaped by a longer history of industrial encroachment, the issue is not just one project. It is accumulated burden. It is memory. It is the afterlife of prior approvals showing up in the body of the present.

This is not only an infrastructure conflict. It is a conflict between worldviews. On one side is an extractive futurism: a worldview that treats land, water, labor, and public systems as inputs for acceleration; that calls depletion innovation when the interface is clean enough; that mistakes scale for destiny; that believes anything can be justified if it can be rendered, financed, and cooled.

That worldview does not only build facilities. It builds consent: by separating the image of futurity from the infrastructures that cool it, and by separating the promise of innovation from the places asked to absorb its costs. This is not merely rhetorical polish. It is an infrastructure of forgetting.

What is being forgotten is structural: the watershed itself, the neighborhood as infrastructure, the public as more than a labor pool or rate base, the history already lodged in the land, the cumulative burden carried before the first announcement. Also forgotten is the fact that reliability for machines is often purchased through unreliability for people.

Black communities already think at the scale of cascading systems. They have had to. Call it survival knowledge. Call it neighborhood intelligence. Call it what happens when people learn, over generations, how one system failure cascades into five more.

Black queer life, especially, has built entire worlds out of relational infrastructure — chosen family, distributed care, safety protocols, quiet communication, resource-sharing under pressure, improvisational redundancy. That is not just social adaptation. That is planning knowledge. That is design theory. That is futurist method, even when no one gives it the grant language.

For artists, writers, designers, and cultural workers, the issue is not simply whether AI can be useful, but what kinds of dependency we are willing to normalize without asking what they require downstream. Every tool trains a habit; every habit builds an infrastructure. Dependency becomes dangerous when it behaves like forgetting.

Afrofuturism, at its sharpest, was never only about being seen in the future. It was always about contesting who gets to define the future's terms — what is valued, what is protected, what is imagined as possible, and what forms of life are treated as collateral.

When water becomes the metric, Afrofuturism is forced back into its most serious work. Not image management. Not corporate prophecy. World-design.

So yes, read the renderings. Read the promises. Read the speeches. But if you want to know what future is actually being built, read the permits. Read the utility filings. Read the watershed. Read who gets redundancy and who gets improvisation. Read what the system loves enough to protect.

What this system requires is not only extraction, but organized forgetting: a way of speaking about the future that severs computation from watershed, growth from burden, and innovation from the lives asked to absorb its costs. In this system, forgetting is not a governance failure. It is a governance technology.

What kind of future requires this much forgetting?