DEV Community

Cover image for The "X Megawatts Powers Y Homes" Trick — and Why "Surplus Power" Has the Same Problem
CarbonLayer
CarbonLayer

Posted on

The "X Megawatts Powers Y Homes" Trick — and Why "Surplus Power" Has the Same Problem

There's a conflation baked into almost every headline energy claim, and once you see it you can't unsee it.

"100 MW powers 85,000 homes" sounds like a statement about capacity — can this resource serve these homes, right now, reliably? It isn't. It's an annual-energy equivalence: total output over a year, divided by average household consumption. Two completely different claims, and the headline always presents the weaker one as if it were the stronger one.

Why does the distinction matter?

Annual-energy equivalence tells you total production balances total consumption somewhere, sometime over a year.
Continuous/instantaneous capacity tells you whether the resource can serve that load simultaneously — including at peak demand, and including during an outage.
A 100 MW resource can match 85,000 homes' worth of annual energy and still be completely unable to serve them at 7pm on a July evening, or during a grid emergency. Nothing about the annual number guarantees the hourly one. Solar and wind make this especially visible — annual output can look enormous while instantaneous availability at any given hour is a coin flip.

The same conflation, wearing a different outfit: "surplus power."

"Surplus power available" gets used to mean at least three different things, rarely disambiguated:

Firm capacity — guaranteed available regardless of conditions.
Expected average output — what the resource produces on average, not guaranteed at any instant.
Residual-after-variable-load — whatever's left over once a primary, variable load has been served first.
These are not interchangeable, and the gap between them is exactly where a technically-accurate claim becomes a misleading one. "We have surplus power" backed by (3) is a fundamentally weaker and more fragile claim than the same sentence backed by (1) — but they read identically in a press release.

Why this matters for AI infrastructure specifically

This is the same accounting-boundary problem CarbonLayer exists to expose in carbon and water claims: a technically true number, reported without the methodology/scope/timing context that makes it meaningful, functions as a false claim in practice. "Powers 85,000 homes" and "runs on 100% renewables" fail in the identical way — both substitute an annual or averaged number for a claim about instantaneous, reliable, simultaneous delivery. Once you're evaluating data center power claims, ask the same three questions every time: Is this annual or instantaneous? Is this firm or average? What's left out of the boundary?

Top comments (0)