Nuclear plants can modulate power by more aggressively cooling the reactor. They don't do this because it's effectively wasting nuclear energy, but it can be done.

There are safety-related limits (power modulation proportion, duration of a pause needed after each modulation, modulations frequency...), and the very combustible status is a major parameter.

Pertinent document (French ahead!): https://new.sfen.org/rgn/expertise-nucleaire-francaise-suivi...

« un réacteur peut varier de 100 % à 20 % de puissance en une demi-heure, et remonter aussi vite après un palier d’au moins deux heures, et ce deux fois par jour »

Proposed translation: "a reactor power output can vary from 100% to 20% in 30 minutes, then after 2 hours can go back to 100% at the same speed, and can cycle this way 2 times per day".

This is quite a good performance when it comes to load-following (French engineers are very good at this), however it is insufficient in the real world (save any ridiculously expensive over-provision of nuclear reactor, most idling) and very weak compared to gas turbines performances.

It can be done, but amounts to decreased utilization, and even greater relative incremental cost, because actual operating cost is the same either way. With nukes already the highest-cost choice, intermittent demand makes them even more impractical.