The House bill to shield households from data center energy costs will be filed under "Washington turns on AI" by most readers. That is the wrong way to read it. The bill does not cap how much power a data center can buy, and it does not cap what it pays. It does one thing: it insists that the cost of serving a very large load sits with that load rather than being smeared across residential bills. For the regulated utilities building into that load, this is not a constraint. It is the political settlement they have been waiting for.
The binding constraint on utility capital plans over the last two years was never physical. Transformers are tight, but they are orderable. Crews are tight, but they are hireable. What actually slows a five year capital plan is a state commission that will not approve it, and commissions have been getting nervous for a simple reason: household bills were rising at the same time headlines said hyperscalers were arriving. A regulator facing an election cycle and an angry ratepayer base has every incentive to trim capex, order a prudence review, or push stranded cost risk back onto the utility. That risk, not demand, is what has kept the market discounting these capital plans.