AI Data Centers vs the Power Grid: Water, Nuclear, and the New Land Rush
The short version: Nelson Barker's video maps the real bottleneck in the AI buildout — not land, but grid capacity, water, and years-long interconnection queues — and it's worth your time.
Our Take
Credit to Nelson Barker for laying out the mechanics plainly: racks climbing from single-digit kilowatts to 70-100 kW under liquid cooling isn't just an engineering footnote, it's the reason substations, transformers, and transmission lines are suddenly the scarcest commodity in America. Northern Virginia didn't become the world's data center capital because of cheap dirt — it happened because utilities said yes, over and over, often with little public scrutiny of what that means for everyone else on the grid.
The interconnection queue numbers cited in the video — roughly 2,600 GW nationally with average waits pushing five years — should stop people cold. That's not a tech story, that's a sign the entire system for deciding who gets power first is broken, and hyperscalers with deep pockets and lobbyists are jumping the line while ordinary ratepayers wait longer and pay more to cover grid upgrades built for someone else's server farm. The water side gets less attention but matters just as much: when a single company's cooling footprint runs into the billions of gallons, that's water your town isn't using for anything else, disclosed years after the fact if at all.
This is exactly the kind of deal-making we track — who signed the interconnection agreement, who approved the water allocation, who's picking up the transmission tab. Check our facility map to see if one of these projects is landing near you, and if you want to push back on secrecy in these approvals, our take-action page has next steps.
This is GridWatch the USA’s original commentary. The video above is the work of Nelson Barker, published on YouTube — full credit to the creator.