The AI story is told in chips. It should be told in dirt.
Every forecast of the next decade assumes compute scales the way it always has — more GPUs, more data centers, more capacity, on command. But capacity isn't conjured. It's built on physical ground, fed by physical power, and right now both are the binding constraint on the entire industry.
Power is the real ceiling
A modern hyperscale campus can draw as much electricity as a small city. The transmission and generation to serve that load take years to permit and build, and interconnection queues across the country are measured in hundreds of gigawatts and the back half of the decade. You cannot out-innovate a line that hasn't been built. The scarcest input in AI isn't silicon — it's a confirmed, bankable path to power.
Power-ready land is scarcer still
It's not enough for power to exist somewhere nearby. A site has to combine real capacity, diverse fiber, buildable acreage, water, workable zoning, and a path through permitting — all at once, in one place. Parcels that check every box are rare, and the moment they're identified they're contested. This is why entitled, powered land in a hyperscaler corridor trades for more than a million dollars an acre while raw ground a few miles away trades for a fraction of that. The premium isn't for the dirt. It's for the transformation.
That transformation is the business
Taking raw or agricultural land in a confirmed-power corridor and carrying it to a de-risked, will-served, entitled position is slow, unglamorous, and relationship-driven. It's also where essentially all the value is created. Most capital wants to own the finished asset. Most developers want to build the building. Very few want to do the hard part in the middle — and that's exactly the part that's scarce.
The discipline that matters now: early and capital-light
You don't need to own the most land to win this cycle. You need to control the right ground, in the right corridors, before the market prices it — and to structure so you capture the value you create without carrying risk you don't need to. Create value early. Keep exposure disciplined. Let the partners built to own these assets for twenty years own them.
The trillion dollars headed for AI is real. Whether it can actually be deployed comes down to a less glamorous question than most of the headlines: is there power, and is there ready ground to put it on?
That's the problem we solve.
Request the Confidential Overview →