Why AI's Biggest Problem Isn't Chips — It's Power!
The short version: MAGNITUD's documentary makes the case that AI's real bottleneck is electricity, not chips, tracing the strain from Abilene's Stargate campus to grid fights nationwide.
Our Take
Credit to MAGNITUD for reframing the AI story where it belongs: not in a GPU shortage, but in substations, transmission queues, and rate cases. The figures they cite — U.S. data centers pulling up to 192 terawatt-hours in 2024, with projections to roughly quadruple by 2030 — aren't abstract. That's demand stacked onto grids that utilities and regulators planned decades ago for a very different America, and someone has to pay for the upgrades, the new gas plants, and the transmission lines being fast-tracked to keep up.
The Texas episode the video flags, regulators hitting pause on new data center connections to assess grid impact, is exactly the kind of moment we watch for. It's a signal that even friendly, low-regulation states are waking up to the fact that unchecked hyperscale buildout can outrun the infrastructure meant to support it. When that happens, the questions that matter are who gets curtailed first during a heat wave, whose rates quietly rise to fund the buildout, and whether the public ever got a real vote on the deal.
This is the pattern we track project by project: the flagship campus gets the ribbon-cutting, and the power and water math shows up later on somebody else's bill. If you want to see how this is playing out near you, check our facility map and find out how to push back through our take-action resources.
This is GridWatch the USA’s original commentary. The video above is the work of MAGNITUD, published on YouTube — full credit to the creator.