A clock on data-centre power

The queue for grid connections used to be simple: first come, first served, and data centres came fastest. The new rule breaks the queue entirely — large computing loads get connected quickly, on one condition: the grid can switch them down when the wind drops.

It is called curtailable load, and it turns the fiercest argument in energy planning — do data centres threaten the transition? — into a timetable. Compute when power is plentiful; pause when it isn't.

Compute that bends

The operators, surprisingly, agreed. Training runs and batch jobs bend well around a schedule; only live services need firm power, and those are a fraction of the load. Two hyperscalers have already signed curtailment contracts covering eighty percent of their regional demand.

The clock matters

Critics note the loopholes: backup diesel is still permitted for “critical continuity,” and nobody has defined critical yet. The regulator promises quarterly publication of every curtailment event — hours, megawatts, reasons — so the public can check whether the clock is real or decorative.

The first scheduled pause happened last Tuesday, for forty minutes, while a front crossed the coast. Nobody's video buffered. Somewhere, a model trained slightly later — and the grid, for once, had one less thing to worry about.