HN Explorer

Units of Economics of LLMs. Reply to Ed Zitron's "AI Is a Money Trap"

6 points by tudorizer ↗ HN
A lot of attention is aimed at the huge investment rounds, cash burned into training foundation models (the trillions mentioned by Sam A.) and not enough analysts explain the units of economics to understand a business.

If you wonder why investors still think it’s a good idea to part with their money, I tried to break down the economic units and long term potential how all this could make sense.

Partial TL;DR

- Cash burn is not a fair approximation for COGS. OpenAI spends mostly on R&D like a pharmaceutical company does. - ChatGPT 4o could be making more than 12.8% in gross margin. - ChatGPT OSS 120B could be making 89% gross margin. It is 90% cheaper than 4o-mini with equivalent reasoning and 3x faster inference. - ChatGPT 5's gross margin is most likely to fall between 12.8% and 89%.

Full breakdown: https://medium.com/@brenoca/openais-road-to-profitability-8c7231f8494b

7 comments

[ 41.0 ms ] story [ 295 ms ] thread
kingkongjaffa
Interesting read!

This stood out to me:

> ChatGPT 5 and ChatGPT OSS are here with the purpose of profitability

This is economically good, but it's also a signal that their capacity to moonshot is stalling either through lack of funding or lack of innovation. They're now pivoting to a more sustainable model.

Models have seen diminishing returns over the last 2 generations of model: GPT3.5 to 4o to 5.

Doubling parameter size does not double model ability/quality.

In the long term models will become commodities that can be interchanged with competitors and open source models, there's no moat, it's not likely anyone is going to sustainably have a hugely better model than the next company.

Claude Code is already showing that you can win in a niche with specialization.

I expect 3 things:

1. We won't see massive jumps on model performance again for a while without new techniques. 2. Model makers will specialize in specific use cases like claude code 3. Moonshot projects like stargate will not have outsized returns, the step change from o3/o4 models to whatever comes next will not be groundbreaking. Partly because of diminishing returns and partly because the average person is bad at explaining what they want an LLM to do.

credit_guy
It's a good analysis, but I am not sure why you are spending time doing this. People who care about your company (investors, users, partners etc) are probably sufficiently familiar with AI to disregard shallow analyses like Ed Zitron's one. You know the saying: a fool can throw a stone in a pond and 100 wise men can't take it out. It's not worth spending time debunking these pieces.
dazamarquez
I would say this article is very shallow. Zitron criticizes what he calls the AI bubble from multiple angles, it's not just "they will never be profitable" — and I agree this would be a wild claim. Even in the worst-case scenario where AI is a giant con, as Zitron paints it, they might just become profitable if they can con enough people. I also don't expect people with a stake in any of this to read Zitron's posts and immediately stop doing what they're doing. That would be silly. I don't think Zitron writes for them, and that what he writes needs "debunking". For how I see it, Zitron mainly advocates for a more critic journalism. Regardless of whether he's right or wrong, he does attempt to critically report on AI.
tudorizer
Right, but critiquing with the right perspective is important. Statements about making a loss must contain the entire economic picture, otherwise they simply aren't true at some point.

A business can't be scrutinized unless the units of economics are understood.

dazamarquez
The piece isn't an independent analysis as the author has an obvious interest in Zitron being wrong. In fact, the piece closes off with a nice marketing self-plug. But that aside, the author doesn't actually refute Zitron's points. One of the main argument is "the comparison with Netflix is wrong", which doesn't prove anything in itself; and then tries to show that inference is profitable. Though just as in their baker analogy, you must factor in all other costs, including training new models. Worthless marketing plug.
tudorizer
Well ... what is independent analysis? The author is not a reporter, but someone who understands business principles.

I think you got the causality the other way around here.