[dgellow]

It's irrelevant... The spending commitment OpenAI has for 2030 is larger than the projected AI infra market in 2030. And we still haven't seen a proof that agents contribute positively to consumers ROI. It's a capital misallocation problem, a number problem, that has nothing to do with subjective feelings regarding LLMs

[simianwords]

In that case, if the numbers do work out and labs make shit load of money, would you agree you were mistaken?

[dgellow]

no... please actually look at the information we know about AI vendors and how much debt they have. OpenAI is projected to have more than $20B in losses for 2026. They wouldn't _just_ need to make a shit ton of money, they would need to first become profitable, then do a shit ton of money while staying profitable, for multiple years, just to match their spending commitment. That's not even taking in account their ROC and company valuation

[simianwords]

Sure but if they do make money and profit, your concern would have been invalid in hindsight wouldn’t it?

Your concern seems to be profitability of OpenAI but that’s easily falsifiable. I’m not saying it’s 100% but I’m saying if they do turn out profitable, your concern wouldn’t have been valid.

[dgellow]

depends on the exact chain of events, but if the economics are indeed solved, and AI labs are found to have a sustainable business model, that concern would be addressed. I would still see the technology as anti-human and would still see the concept of an agentic economy as deeply unserious, wasteful, and risky. Not exactly sure what your point is though.

FWIW my concerns aren't only for OpenAI, it's just the poster-child of the AI bubble. SpaceX AI strategy and investment is a complete joke (the S-1 they filled is an insult to a reader's intelligence). Anthropic seems to have been more cautious but has the same fundamental economic problems. Nobody in that whole industry has a moat, the top AI vendors are way too exposed to Chinese/open-weights labs. The datacenters debt investment vehicles (SPVs) look extremely shady and made to obfuscate the underlying assets. The HBM manufacturers seem to be going through one of the most violent boom-burst cycle ever. Oracle is... doing Oracle things, I would be shocked if that company is still alive in its current form in 5y. And so on

[simianwords]

Here’s what you said earlier

> It's a capital misallocation problem, a number problem, that has nothing to do with subjective feelings regarding LLMs

Now you say

> I would still see the technology as anti-human and would still see the concept of an agentic economy as deeply unserious, wasteful, and risky. Not exactly sure what your point is though.

You can see how someone might be confused with this. Is it your subjective feeling? Or objective economics? I’m not interested in your subjective feelings on AI.

On objective economics: your concern would be invalid if the AI companies overall make profit. If they don’t, then you would have been right on the point of being careful.

[dgellow]

No, I don’t see the confusion. There are a lot of distinct issues with LLMs. In my initial message I made it clear I was talking about the economics issues.

> your concern would be invalid if the AI companies overall make profit. If they don’t, then you would have been right on the point of being careful.

No, that’s fallacious reasoning. If the AI vendors are currently profitable that doesn’t mean the concerns regarding their unsustainability is invalid, but that changes the calculus depending on the details. They could be profitable right now on paper and still be economically not viable. The core problem is that the amount of money allocated to the AI bet is completely disproportionate compared to the actual economical value of the technology. That’s where the imbalance is. The exact reasons for the imbalance can change over time, but there would still be a misallocation. The companies would need to be profitable, and/or their expenditures would need to pay off, and the overall demand needs to grow exponentially, and they need to be protected from Chinese competitive pressure, etc.

I still don’t understand your overall point