Re:I'm Not Convinced the AI Bubble Exists
by PinkaDunka on Monday November 09, 2026 @08:03AM
(#50425023)
[delayed]
Re:I'm Not Convinced the AI Bubble Exists
by Traster on Monday November 09, 2026 @08:21AM
(#50425234)
It means that the companies have to be highly cost-competitive, they can't win market share and then raise margins. If (as is the case today) astra is bad value for money compared to opus-5-5, then I'm switching, and if (as is the case today) for lots of stuff gemini can do the low hanging fruit, I'm switching. And all of this is happening before my company has really put any boundaries on our spend, in the future I'm going to be on a budget and that's really going to impact my decisions.
Re:I'm Not Convinced the AI Bubble Exists
by PinkaDunka on Monday November 09, 2026 @08:36AM
(#50425411)
But according to Anthropic they are getting the healthy margin right now. If they raise margin it would be “even better”, not just “finally making any money”
Re:I'm Not Convinced the AI Bubble Exists
by Traster on Monday November 09, 2026 @08:57AM
(#50425682)
But it's not clear that's sustainable. If you look at what happened in the first half of the year, they were very clearly ahead and that translated into both massive growth and great margins. But that looks like it's atleast slowed or maybe even reversed in Q3. And again, that's while there's a shortage of compute and so everyone who has compute can charge a premium for it. Shoot forward 1-2 years as this massive glut of compute hits the market, can Anthropic sustain their margins or do they have to compete with Meta and Alphabet dumping almost free inference to claw back costs on data centre investment.
Depends if the vibe changes, right now people are switching based on capability and not cost (unless it's entirely unreasonable like release Fable).
If we start to see the main switching factor be cost or a plateau in capability then I would be worried.
>In theory, an AI company has to be worried about their users switching over to a competitor or using an open source model instead. In practice, this isn't a major concern. For starters, the average person...
There's two things to disagree with here. First, this isn't theory, I work at a company and use AI heavily, switching model is as simple as changing a line in a json file. And would I do it? I've done it like 5 times this year. Claude was good, then openai was for a while, then there were times gemini was great for quick trivial tasks, now I'm mainly back on opus-5-5 with astra for some more complex tasks. Secondly, I'm not the average person, no, I'm the person openai and anthropic actually make money from. Sure, a consumer might be stickier, but so far they're not really making money from consumers, they're mainly doing B2B.
The core thing that bugs me is not so much that AI is going to be massively unprofitable, I think there's definitely going to be profitable companies. I just think if you're Google, the best case scenario is you're transitioning from fantastic economies of scale in a monopoly business to a business with costs roughly proportional to revenue and no moat. That is really bad for Google and it's not unreasonable to think it's motivation enough for them to try and undercut Anthropic and OpenAI until they're bankrupt.