Do remember that we're discussing a case of illuminating Earth (and only Earth) with some dispersionless, cylinder-like, 100% efficient laser beam with perfect accuracy. Even then, with those idealisations, power scales with the square of the radius of the thing we want to illuminate. Accuracy is also fun: in our case, it's like pinpointing something roughly the size of a credit card on the surface of our Moon, but without the joys of 4.3 years worth of one-way delay or tracking a moving object.

Also, I didn't say that divergence isn't significant. Just that it likely won't involve higher maths to find an approximation, which is semi-true. Had to do a double integral over a disk to get from intensity [W/m²] to power [W].

Here's how we can calculate the power delivered by a Gaussian beam, and it's ripe for plugging numbers in. I took the formulae and symbols from the article. There's also a calculation of how narrow the beam would have to be at its narrowest point, which turned out to be essentially zero (which I, perhaps mistakingly, interpreted as equivalent to a point source). Pinging am_Unition for peer review and help in moving it forward. It's not pretty, though. My initial intensity assumption goes asymptotically to infinity the narrower the beam, so there's possibly a problem/fuckup.

I absolutely encourage everyone to play around with the numbers. Maybe it could work for other wavelengths?