As artificial intelligence companies race to secure enormous amounts of land, electricity and water, a new proposal in Congress is drawing a sharp line around one resource: federal public land.
The proposed legislation would permanently prohibit qualifying large AI data centers from being built or operated on land owned or managed by the federal government, including certain military installations. If enacted, the measure would represent one of the strongest federal restrictions yet on where the infrastructure behind America’s rapidly expanding AI industry can operate.
And the proposal would reach far beyond the data center buildings themselves.
It would also restrict supporting infrastructure associated with qualifying AI facilities, including electrical substations, transmission equipment, natural-gas infrastructure and large-scale water-cooling systems.
That matters because the physical footprint of AI is much larger than the server racks most Americans picture when they hear the term “artificial intelligence.”
AI’s Growing Appetite for Power and Water
Modern AI models depend on massive computing facilities packed with specialized processors. Keeping those machines running requires substantial electricity, while keeping them from overheating can require sophisticated cooling systems and, depending on their design, significant quantities of water.
As demand for AI services grows, technology companies and data-center developers are looking across the country for locations where they can obtain reliable power, sufficient acreage, transmission capacity and access to cooling resources.
Federal land could therefore become attractive for certain infrastructure projects.
The proposed bill effectively argues that such land should be off limits.
Supporters contend that federally controlled property exists for broader public purposes—including conservation, recreation, national defense and the interests of future generations—and should not become a convenient source of land for privately owned AI infrastructure.
Existing Facilities Would Not Simply Be Grandfathered In
One of the most consequential elements of the proposal is its treatment of existing projects.
Rather than allowing facilities already operating or under construction to remain indefinitely, the legislation calls for federal agencies to begin acting shortly after enactment.
Within 30 days of the legislation becoming law, affected agencies would be required to begin shutting down operating facilities, halt qualifying projects under construction and start the process of removing covered infrastructure.
The land would subsequently have to undergo cleanup and restoration.
That provision makes the proposal considerably more aggressive than a simple prohibition on future construction. It could potentially affect investments that have already been made, raising questions about contracts, stranded infrastructure, implementation costs and possible legal disputes.
Those details deserve attention because “banning future projects” and forcing existing projects to close are economically and legally very different policies.
The Case for Protecting Public Land
Supporters of the proposal are likely to frame the debate around a straightforward principle: public land should primarily serve the public interest, not the infrastructure needs of private technology corporations.
From that perspective, the AI boom should not automatically override longstanding concerns about land conservation, wildlife habitat, recreation or water availability.
There is also a broader question of precedent.
If the federal government allows large AI facilities onto public property because they are considered strategically important today, what other energy-intensive private industries should receive similar access tomorrow?
For communities already concerned about electricity prices, groundwater supplies and rapid industrial development, that question could resonate well beyond traditional environmental politics.
But There Is Another Side to the Debate
A nationwide prohibition could also carry significant costs.
The United States is competing with China and other countries for leadership in advanced artificial intelligence. Building that computing capacity requires physical infrastructure somewhere, and restricting potential locations could increase development costs or delay new projects.
Critics could therefore argue that an absolute ban is too rigid.
They may contend that instead of prohibiting AI facilities on all federal land, Washington could establish case-by-case environmental reviews, strict water-use limits, electricity requirements, lease payments and restoration obligations.
There is also an important distinction that can easily disappear in political arguments over data centers: not every facility consumes the same amount of electricity or water. Cooling technology, power sources, facility size and geographic location can dramatically change the environmental footprint.
Treating every project identically could therefore sacrifice potentially beneficial developments along with genuinely problematic ones.
The Bigger Fight Is Just Beginning
The legislation has not passed Congress and is not currently U.S. law. Its introduction does not mean that large AI data centers are about to disappear from federal property.
But the proposal highlights a conflict Washington will increasingly have to confront.
Artificial intelligence may appear to exist in the digital world, but its foundations are unmistakably physical. AI requires servers, power plants, transmission lines, cooling systems, water and land.
As America’s AI industry expands, policymakers will have to decide not merely whether the technology should grow, but where the enormous infrastructure supporting that growth belongs—and who should bear its costs.
The central question is therefore bigger than any single data center:
Should federally owned land ever be used to support privately controlled AI infrastructure if doing so advances America’s technological competitiveness, or should public land remain permanently protected from that kind of commercial development?
How Congress eventually answers that question could help determine not only the future of America’s public lands, but also the physical shape of the country’s AI revolution.