If you enjoy this free newsletter, why not subscribe to Where's Your Ed At Premium? It's $7 a month or $70 a year, and helps support me putting out these giant free newsletters!
At the end of November, NVIDIA put out an internal memo (that was leaked to Barron's reporter
I have a little perspective to offer on this:
One of the core points of the author’s argument is, “Generative AI providers are operating at a loss, and I don’t see how they could become profitable”, specifically calling out ‘where are the new customers’?
I work at a company that is a direct customer of OpenAI and we get real, useful output from it. If they were to double or even triple the price, I’m almost certain that we’d just pay it (as it’s I very cheap at the moment). I’ll admit I’m not actually running numbers, iirc several of the companies mentioned were near break even, so doubling revenue would make them very profitable.
So my argument here is that I don’t think they need new customers, they just need to raise prices. Launching with unprofitable prices and jacking them up later once people get a taste is part of the standard tech industry playbook and wouldn’t be the least bit surprising in this situation. The company I work for has only small integrations so far, but I can imagine that at some companies almost their entire service is based on using LLMs, in which case the providers would literally have the ability to charge whatever they want.
Also fwiw, I’m a gen ai hater, and wish it would collapse and go away, but I’m not entirely convinced by this guy’s argument, but maybe if I looked harder at the numbers it might become obvious that the price increases I’m suggesting still won’t cover investment expenditures.
It’s kind of the same when any new tech enters the market, first you get your product out there below any competitors to capture as much of the market as possible, then you slowly raise prices.
Some may be closer to break even, but openAI is tens of billions in the hole (something like 13.5 billion net loss in the first half of the year alone).
Sure, they could triple prices but then you lose a percentage of customers etc etc.
Yeah, I did realize that the break even ones I think largely are using OpenAI as part or all of their capacity, so they probably aren’t really break even for that reason - OpenAI is subsidizing them.
I appreciate how this newsletter/article does a solid job of highlighting some of the weird dealings we may be seeing that make these companies look more viable than they are (very much including these companies essentially subsidizing each other).
The big thing I keep coming back to is that there doesn’t seem to be a path to profitability for the LLMs. Chapo Trap House had an interview with Ed Zitron and he highlighted that some of the power users are pulling down over $50k in generation/prompt costs while spending $20 a month (or whatever, I’m spitballing the numbers).