The grid nobody planned for: public power, hyperscalers and the race to rewire America for the AI age

The US went from essentially zero load growth for twenty years to 3% national growth almost overnight. The supply chains, permitting pipelines, engineering workforce and regulatory processes were all calibrated for a different world. Bridget van Dorsten is joined by Tom Falcone, President of the Large Public Power Council, representing the 30 largest publicly owned utilities in the United States, collectively owning around 85% of public power assets and currently serving roughly 18% of all US data centre load.

Tom explains what makes public power structurally different from investor-owned utilities: locally governed, not-for-profit, and built to minimise cost rather than earn a return on equity. That governance model turns out to matter a great deal when trillion-dollar hyperscalers come looking for power. Public power utilities have no financial incentive to favour their own assets over a customer's, and their local accountability makes deal-making faster and more direct.

Bridget and Tom also work through the mechanics of how the industry is actually responding. Large-load tariffs are reshaping the interconnection queue, forcing hyperscalers to make long-term financial commitments rather than reserving capacity for free. About two thirds of speculative requests disappear once real commitments are required, which tells you something about the gap between announced demand and real demand. LPPC members are nonetheless planning to add around 60GW of new generation over the next ten years to meet load that is forecast to grow from 4GW to 18GW of data centres in their territories alone, in just five years.

The episode also tackles private use rules, a Treasury regulation from 25 years ago that nobody expected to become a bottleneck for the AI era, the capacity factor realities that make peak-day power so much harder to deliver than annual energy, the nuclear question and why federal involvement is probably unavoidable if the US wants to build at scale, and where CCS can and cannot realistically be deployed.