Span's 'deflationary' pitch just shifts the bill to homeowners

A distributed load is being marketed as a cheaper one, and the difference matters for who ends up holding the bill.

Goldman Sachs projects data centers will push U.S. electric bills up about 6% over the next year. The Houston Advanced Research Center forecasts 399 billion gallons of water drawn in Texas alone by 2030. McKinsey sees $7 trillion in AI capex by 2030. The appetite for a counter-narrative is real, and Span's Ryan Harris is pitching one: gigawatts of residential compute, Harris says, could land "in a deflationary-to-energy-price way," through cabinet-sized XFRA nodes installed on PulteGroup homes and small businesses.

The mechanism on offer is not a price cut. It is a load-shape bet (a characterization reporters interpret as): the same compute that would have sat on a hyperscaler's substation gets sliced across thousands of residential ratepayers, each absorbing a slice of the noise, the thermal load, and the interconnects. The unit cost of the compute does not fall; the political cost of hosting it relocates.

Distributing the boxes does not retire the bill. It moves the protest sign from the data center county to the cul-de-sac.