Brief notes after spending 10 minutes figuring out how to get this, not nitpicking or saying you got anything wrong. Source was page with "54" at the bottom
- Unvested grants over 4 years vs. buybacks this year
- #s are for 2020, not 2023
- # excludes tax witholdings (i.e. shares they'll never grant).
- For 2020, grants of 15.539 - 10.273 witholding = ~5B, matching sibling comment's correction
- $2.324B is one of 4 columns, real number is $50.274 B
- so for 2020, real #s are 5.26B in unvested grants over 4 years versus 50.274B in buybacks that year.
- for 2023, 22.578 - 10.164 = 12.4B in unvested grants over 4 years, versus 62B in buybacks that year
You're right, it's too low, see link at bottom: my guess is the $2B number, inclusive of all granted but not vested, is the correct #.
In general people way overestimate the # of SWEs at Google relative to the employee #, even within Google.
I don't wanna get too cute in public but some fermi estimates including, say, number of employees per engineer at a 100 person startup, and a jaundiced yet correct view of how much is outsourced at a company with a lot of grunt level SWE work common to every company, bridges the multiplier from there.
Poster is correct: that is not the reason for them.
We can confirm this. Compare RSUs granted/year -- $150M -- to buybacks this quarter -- $70B. ~700x as much.