AI Data Centers Claim 24,000 Acres of US Public Land
Developers filed 35 applications to build AI data centers and infrastructure on federal land across six Western states.
Policy · Source: Reuters (via SRN News)
What happened
The AI infrastructure boom is spilling directly onto federal property. Developers have filed at least 35 applications to build data centers and related infrastructure on US public lands. These active and pending requests cover more than 24,000 acres across six Western states. The targeted states include Arizona, Idaho, Nevada, Oregon, Utah, and Wyoming.
The Center for Western Priorities analyzed records from the US Bureau of Land Management to uncover the scale of this land grab. They found proposals for 17 massive data center campuses. Another 13 applications cover the necessary support structures. These include transmission lines, water pipelines, telecommunications facilities, and access roads required to keep the servers running.
Specific developers are already staking their claims on the map. Prometheus Hyperscale has applications tied to a facility in Wyoming. Copia Power is pushing a range of projects across Arizona, Nevada, and Idaho. The Trump administration strongly backs this expansion as a national priority. They are actively seeking to open public lands to a wider range of industrial uses, including technology and energy projects.
Key facts
- 35 — Pending or active data center-related applications on federal land
- 24,000 — Acres of US federal land targeted by developers
- 17 — Proposed data center campuses across six Western states
- 13 — Related infrastructure projects like pipelines and transmission lines
Why it matters
Builders need to understand that the compute bottleneck is no longer just about securing chips. It is about finding the physical space, power, and water to run them at scale. Moving onto federal land shows exactly how desperate hyperscalers and developers are for real estate. Commercial land with adequate grid access is drying up. Federal land offers a new frontier for the massive AI clusters required to train next-generation models.
This shift sets up a brutal collision between tech companies and local communities. The public is increasingly opposed to these massive infrastructure projects. Locals are citing severe concerns about electricity demand, water consumption, wildlife impacts, and the loss of public access to federal lands. If public opposition and environmental reviews stall these federal land approvals, the supply of new data centers will tighten. That means the cost of hosting and training large models will remain a heavy burden for startups.
For builders
Compute costs hinge on land approvals
Building on federal land requires navigating heavy bureaucracy and local pushback over water and wildlife. Founders relying on cheap future compute should expect delays as data center developers fight for these permits. Do not assume server costs will drop overnight.
Opportunity in resource efficiency
Data centers face massive public backlash over their intense water and electricity consumption. Startups that optimize model training efficiency or reduce hardware power draw have a massive advantage. You now have a direct pitch to hyperscalers trying to appease angry local governments.
My take
The AI race is no longer just a software problem. It is a massive real estate and energy land grab. If we are seriously planning to pave over public lands and drain local water supplies to train the next frontier model, we need to ask if our current compute architecture is fundamentally broken. Builders need to focus on efficiency before the public shuts off the power.
Original reporting: Reuters (via SRN News). This is my rewrite and opinion.