More like 25 million, but sure.
No, it is most definitely
not 25 million or 'half a Cherry.' If it was half a Cherry MX, there wouldn't be a single working M left.
I took a number of M's and my own logged keystrokes and extrapolated the data with a known value keyboard - specifically a 1988 Model M that had been in continuous use for a minimum 8 hours per day, 100WPM, since 1988. (I in fact, knew the person who used it, and bought it used from their employer.) It was used for medical transcription.
100WPM for 8 hours == 360,000 words == 2,160,000 (that's 2.16M) keystrokes per day.
Let's cut that to just 150,000 words - that's still 900,000 keystrokes a day. In other words, 25 million operations? Would not have lasted a year. Which works out to about 297,000,000 operations divide by 37 keys gives you 8,027,027 ops/key average. Which by the way, isn't an atypical use for a Model M. At all.
Now expand that to 10
years of continuous operation. That would give you a total of 2,970,000,000 operations (or 2.9 BILLION) with an individual key receiving an average of - let's round down to 8,027,000 for sake of argument - 80,270,000 operations per key. That's 80
million meaning that even accounting for unequal dispersion it is impossible that any given key did NOT receive well in excess of 50 million operations - notably, the space bar!. When I purchased it, the keycaps were worn to hell, but springs still functioned like new. 25M operations is hilarious sandbagging at it's best, and downright disingenuous at it's worst. I have a process control keyboard (3151 base, 1987) here which was used to run the same set of keys hundreds and hundreds of times a day. There's a bit of mush, the caps look awful from grime, but it
still works flawlessly.
These things were purchased and used because they are virtually indestructible. Not because they are less reliable than the competing Cherry MX and not because they were cheaper. People bought IBM Model M's because they wanted a keyboard which was more reliable, and they were prepared to pay a premium to ensure no cheap foam-over-foil failures or abruptly faulty Cherry MX based keyboards. (Now you know why Black was popular back then.) These are designed to stand up to constant use and abuse in environments most people wouldn't dare - steel plants, iron plants, mining facilities, medical facilities, the list goes on. They get beat to crap, abused, and thrashed. And we're still buying them 20+ years later, cleaning 'em off, fixing a few minor spring breakages due to bad removal and bad capacitor choice, and what you will end up with is another keyboard that will last under those conditions for well over 10 years. I bought it in 1998. I used it until 2009sh? It never failed me.
So yes. 2 billion keystrokes out of a Model M keyboard is a very reasonable and predictable number. Individual components may fail, but within repair scope. Actually rendering it dead requires a great deal of work or bad luck.