With basically almost nothing coming online for decades and the handful of projects that do come online being late, hundreds of percent over budget, and delivering only a relatively tiny amount of power (relative to the equivalent of hundreds of new reactors worth of solar/wind/battery every year), it's hard to maintain that nuclear has any role whatsoever to play. No amount of magical thinking is going to change that any time soon.

The AI revolution will be mostly over by the time any substantial amount of new reactors come online. Whatever power they'll use, it will mostly not be nuclear. It's a lot of gas right now that will likely shift to a mix of much cheaper solar and wind. Nuclear powered AI will be a rounding error.

Maybe somebody will figure out how to do new nuclear plants in under a decade (good luck!). That would be spectacularly fast. But if it's not planned and approved right now (and very little is), we're basically talking the 1940's for any significant new capacity to come online. That would still be a tiny fraction of the yearly growth in renewables. That probably will accumulate to about three orders of magnitude more power generation in the time until then. With hundreds of GW added per year, we are talking at least a few TW of generation cumulatively.