alt Hacker NewsBut how does that detract from the overall point that Open AI is losing billions of dollars? If there's an insatiable demand for AI compute, where's the profit?
Replies
High growth companies often have significant negative cashflow during the early high growth era, followed by positive cashflow in the years later down the line.
This phenomenon is known as the J-curve[1], and Uber is a good example of how this can turn out absolutely fine. To some extent, the entire Venture Capital industry exists to finance precisely this dynamic!
Nb. I'm not suggesting OpenAI is fairly valued, or that they will definitely become profitable, but "OpenAI is losing billions of dollars" doesn't really mean anything in and of itself.
[1] https://www.uark.vc/blog/breaking-down-the-j-curve-the-journ... (many other similar such articles exist)
First and foremost, it's about competent reporting. You can think a company is doomed and still expect people to report on it accurately.
Zitron continually presents things in ways that create a hyperbolic narrative. Turning routine consolidation accounting into an unexplained mystery hinting at some sort of fraud is the perfect example of that. He does it so often and in such a way that I truly believe he just doesn't understand accounting.
It doesn't take a rocket scientist to understand that OpenAI is losing money. But the question ("where's the profit?") assumes that profit is the thing being optimized for. It isn't.
There are basically three buckets here: cost of serving a query, cost of training and capex for capacity.
The first one is unit economics. The other two are bets on the future that get expensed against present revenue. A company can serve every query at a healthy gross margin (OpenAI has improved margins considerably) and still have a $9 billion loss because it spent $12 billion training a model that generates $0 this year.
Zitron constantly blends everything into a pithy "they lose money on everything" narrative, which just isn't accurate. The thing is that OpenAI could have a very different P&L if it chose to, say, stop training the next model.
The problem, obviously, is that if you stop training the next model, the competition might eat you. So right now you have a situation where the the frontier model you spent $12 billion on depreciates in about 18 months, the GPUs depreciate on a schedule nobody agrees on, and you seemingly can't stop the cycle without risking your position in the market.
This is the legitimate bear case, but the problem with Zitron is that he doesn't make it using an argument that is coherent and honest as far as the accounting is concerned. And the accounting is everything.