Hacker News new | ask | show | jobs
by artninja1988 36 days ago
Has anyone done the math on how many jobs AI has to replace in order to justify current valuations? Would be interesting to see imo
3 comments

There's no public valuation of any AI company.

But by the investment contracts OpenAI announced, if they replaced every white collar job in the world and captured all of their salary, they would have a P/E of ~1/70. (Assuming no costs at all.)

You decide what return horizon you are comfortable with and solve the linear equation for it. (But my numbers are from the end of last year, I haven't seen any more recent compilation of their contracts.)

Could I please ask what stats you used? Intuition says the valuation is too large and implies replacement, would be interesting to see something backing it up.
The 1.5T investment size was widely reported last year. AFAIK, they are required to spend all that money in the next few years if they survive for that long. Here's one place reporting it:

https://www.business-standard.com/world-news/sam-altman-open...

OpenAI also announced enough deals this year to more than double that amount. But I don't think anybody has put those announcements together and deduplicated them.

That part I get, the white collar work part is what I'm looking for.

Thanks for the response.

I think the goal is to replace workers who cost $10,000/mo with an AI agent that costs $8,000/mo
$8000/month is way too low. If you're only spending $8000/month on AI you're in drastic danger of being replaced by a more efficient AI firm.
You should replace a team of 7 costing $70,000/mo with a team of 2 costing $30,000/mo and $100,000/mo in ai tokens.
If you replace 100 then that's 800k/mo spend. Would that do?
That's a decent start, but if we're talking shop you might as well bump to at least 1.2-1.3M just to play it safe.
My back of the napkin was a floor of five to ten percent just to break even with the physical buildout + a couple of years of energy bills.