The Long Run
Thesis

The Department of Energy reports that US data centres consumed approximately 4.4% of total US electricity in 2023. DOE and Lawrence Berkeley National Laboratory modelling estimates that figure could reach approximately 6.7% to 12% by 2028.

The precise outcome is genuinely uncertain and we are not going to pretend otherwise. What is not in dispute is the direction, and that those megawatts have to come from somewhere.

Berkeley Lab's 2026 Queued Up report finds roughly 8,200 active projects seeking transmission interconnection at the end of 2025, representing 1,312 GW of generation and approximately 749 GW of storage — more than 2,060 GW combined. For projects that entered commercial operation during 2025, the median duration from interconnection request to commercial operation exceeded five years in regions with available data.

That mismatch is the opening. It creates value for capacity that can be deployed or coordinated using infrastructure that already exists — which is a different question from whether any particular company captures that value.

The number that matters may ultimately not be how many solar panels get installed. It may be how many controllable energy assets sit behind those installations.

The shift is visible in the reported numbers. Storage attachment reached 74% in Q2 2026, against 70% in the prior-year period, and the networked fleet now represents approximately 4.6 GWh.

In July 2026 Sunrun announced that its California distributed power plant had expanded to more than 80,000 participating households and over 110,000 home batteries, providing up to 425 MW of peak dispatchable capacity. The programme launched in 2024 with approximately 16,000 enrolled customers.

California is the largest example but not the only one. Sunrun has reported that more than 100,000 customers participated across 18 distributed power plant programmes during 2025, dispatching nearly 18 GWh into grids over the year.

So the useful question was never whether distributed power plants might eventually work. It is how valuable they can become, and that is an economics question this thesis does not yet have an answer to.

The next two sections take them one at a time, because collapsing them into a single bullish sentence is exactly the error this project exists to avoid.

Stated accurately, it is still interesting. The proposition is that rather than waiting exclusively for new power stations and transmission, existing behind-the-meter resources could be coordinated at scale to release and shift capacity. That is a real idea being taken seriously by serious operators.

An announced commercial arrangement is a materially stronger signal than a non-binding framework. It is the clearest evidence to date that the distributed fleet is beginning to seek customers beyond the homeowner.

That is a pilot. Not a product line, not a disclosed revenue stream, and not evidence that a residential solar company has become a compute provider.

Read that table honestly and it says two things at once. The installed base is still growing and the mix is shifting toward storage — both supportive. The rate of new additions and the volume being installed are falling sharply — both damaging. A thesis built on a compounding network cannot wave away the second half.

Several of these can be true at the same time as the thesis. A number of them can be true and still leave it directionally right but untradeable, which is its own kind of wrong.