Ask HN: Does OpenAI make a profit on GPT-3.5-turbo API?

[steveridout]

Like many, I'm using the OpenAI API in my product and I've noticed that gpt-3.5-turbo offers very good value for money relative to the other models.

My main concern is that OpenAI might be running this at a loss, making a price increase likely.

Does anyone here have any educated guesses about their likely unit economics?

[tikkun]

> My main concern is that OpenAI might be running this at a loss, making a price increase likely.

Their goal is explicitly to only reduce the prices of all of their models over time. If the increase the prices of any of their models, something would've gone very wrong or they would've made a major strategy change. I've read pretty much everything OpenAI's published and talked about, and they've been consistent in mentioning they want to continually reduce the price of models.

[steveridout → tikkun]

Thanks. I did an (admittedly quick) search to confirm this but didn't find anything. Do you have a reference by any chance?

[tikkun → steveridout]

I wish I could find the quote. It was something like:

"You gotta assume we're going to keep making the models better and faster and cheaper." (My memory of an old interview)

There was also this from a recently deleted blog post:

"Cheaper and faster GPT-4 — This is their top priority. In general, OpenAI’s aim is to drive “the cost of intelligence” down as far as possible and so they will work hard to continue to reduce the cost of the APIs over time." (Now deleted blog post from earlier this week)

[huevosabio]

From a recent podcast [0] it seems unlikely that they make profit from gpt-3.5-turbo, at least not in the sense that it pays for employee salaries. Instead, from the conversation, it seems like OpenAI is aiming to make money mostly from enterprise partnerships, probably to fine tune GPT-4 or what not to your own private data.

I think this makes sense, you keep competitors at bay by offering a good-enough, dirt-cheap API and then ask the big bucks for the private setup. This mirrors Stability AI's strategy, which is basically that but with open source models.

[0] https://open.spotify.com/episode/2v8Sa1RNnfZa6l7JIYmIL2?si=o...

[neximo64]

Yes. and that is because the Boom 176B model isn't crazy to run on. Infact if there are enough clients theyd get economies of scale.