When can a power company take your land for data center infrastructure?

Once that was cleared up, the conversation landed on a harder question. Power lines are the classic case for eminent domain, but many people do not buy the idea that a line built mainly to serve one giant private data center should automatically inherit the same public-use legitimacy as a highway or municipal water line. Others pushed back that the grid does not work customer by customer. New transmission is shared infrastructure by definition, and blocking it because the initial anchor load is a data center just turns every expansion into veto-by-adjacent-landowner. A lot of the practical detail came from people familiar with utility takings. In most cases the owner loses an easement, not the whole parcel. Compensation is often above fair market value because developers want to avoid condemnation fights, though several people argued that standard fair market value still underpays because it ignores the owner’s reservation price and the permanent nuisance of towers, access roads, noise, and use restrictions. That compensation point became the cleanest middle ground. Eminent domain may be necessary for linear infrastructure, but it should be expensive enough that utilities and governments only use it when the project is genuinely worth forcing through. The other big thread was whether these loads should move to power or power should move to loads. Some argued that data centers, especially training-heavy AI facilities, could sit nearer cheap generation in the wind belt and let fiber do the long haul. Others said that only works for some workloads, ignores labor and network economics, and misses the broader reality that the grid already needs long-distance transmission to move renewable power to cities and industry. In that framing, data centers are not the reason to build a modernized grid. They are just the latest stress test exposing how hard the US has made any serious transmission project.