It's not literally impossible, of course. But it's super unlikely. A 2.5m December has only been achieved twice, once by PS2 and once by Wii. Switch is not as popular as either of those machines.
Beating PS4 for the year wouldn't require a 2.5m December, though, and is much more likely.
Roughly, current 3DS + Switch revenue in the Americas is about 2.25x what 3DS+Wii U revenue was (approximately) launch-aligned. It's about half what DS+Wii revenue was in a comparable timeframe. Relative margins are bigger now (undoubtedly drive by high-margin amiibo, mobile, and eShop earnings).
The gap is ~666k, not 800k as has been said a couple times in the thread.
I could mention that Let's Go absolutely does not have sales "nowhere near where [Pokemon] usually is". Or that, yes, having the highest-selling exclusive game of all time could pretty much make a year just by itself (especially since Switch will only pass PS4 in the month of that game's release). But there's a larger point to make besides these errors.
Which is that no matter what the reasons for it, massive hardware numbers in year 5 are historically more difficult than massive numbers in year 2. Looking at the past two decades, there have been 3 year-two totals that meet or exceed what Switch will do. That's out of 9 machines (PS2, Xbox, Gamecube, 360, PS3, Wii, WiiU, XB1, PS4), or 33%. But how many fifth years were above PS4's 2018? That's only 1 (the 360), or about 11%.*
*If you add handhelds, then both Switch year two and PS4 year five are beaten by one more entry each (GBA and DS, respectively).