Here's the link to the original source:
Since it is a Mat Piscatella writeup, it's to be expected that people end up confused on which percentages relate to what, because raw numbers are repeatedly obfuscated in typical NPD style.
55m mobile gamers have played Pokémon Go at some point. Of those 55m, 60% are still active players, so that results in a base value of 33m for further statements.
24% of active Go players also play games on Switch (~8m). Since this value exceeds the LTD sales of Switch in the USA, the research doesn't make a distinction between Switch ownership and playing Switch, so any given Switch console can have multiple players. Note that Piscatella does not mention Switch ownership either. 65% of these 8m intend to purchase Let's Go, so that's ~5.3m. Piscatella's doublecheck is that this number (5.3m) equates to 15% of active Go players; 15% of 33m is ~5m, the small discrepancy can be explained with rounded percentages.
13% of active Go players who do not play games on Switch (meaning 13% of 25m, because 8m of 33m do play games on Switch) intend to buy and play Let's Go. That's ~3.25m. What's noteworthy here is that this stat does not equate to purchase intent for Switch, because these people may or may not already have a Switch in their household, but it's owned and played by another member in the household.
This research suggests that ~8.5m people in the USA have a purchase intent for Let's Go, but intent doesn't mean that they'll go through with it. The research doesn't reek of incompetence, because sales expectations for Let's Go in the USA should be at least half of that purchase intent figure. I've seen far worse.
Bottom line is that the thread title as well as the numbers in the OP are inaccurate. There aren't 5m Go players who plan to buy a Switch for Let's Go. The actual figure is ~3.25m and comes with the caveat that a Switch purchase is not mandatory in every realistic scenario. As for the question if Let's Go will sell millions of copies and shift Switch hardware, that was never in doubt to begin with.