You may have noticed I left something out in that last post about Andreesen and Horowitz — their political vision is focused on crypto, AI, and a tax policy they like better
. I said nothing about AI! You may praise me for my restraint.
I shall now correct my omission. Here’s the breakdown of the money OpenAI is spending on this boondoggle:
Total revenue has been $283 million per month, or $3.5 to $4.5 billion a year. This would leave a $5 billion shortfall.
Training the AI models will cost OpenAI about $7 billion in 2024. For ChatGPT alone, the cost will be $4 billion. New models may add $3 billion to that cost — OpenAI has had to train new AI models faster than it had anticipated.
Microsoft’s OpenAI “funding” is largely in the form of Azure compute credits. OpenAI gets a heavily discounted rate of $1.30 per A100 server per hour. OpenAI has 350,000 such servers, with 290,000 of those used just for ChatGPT. This cloud estate is apparently running near full capacity.
Staffing costs for 2024 are likely to be $1.5 billion, up from $500 million for 2023. The median OpenAI engineer salary in 2023 was a $300,000 base salary and $625,000 of stock-equivalent compensation. [Bay Area Inno, 2023; Levels.fyi]
They’re spending roughly twice what they’re making. Almost all their servers are chewing away on ChatGPT. And personally, the worst of all as far as I’m concerned is that software engineers are getting paid $300,000 base salary and $625,000 of stock-equivalent compensation. If my employer paid me half that amount of base salary, rather than a quarter, and never mind the big stock bonus, I’d be coasting on easy street and could hire a live-in masseuse and, I don’t know, go crazy and buy a second car? I struggle to imagine that much money.
I suppose my daughter would have the qualifications to get into that kind of business, but I’d encourage her to keep her soul intact.