●Stories
●Firehose
●All
●Popular
●Polls
●Software
●Thought Leadership
Submit
●
Login
●or
●
Sign up
●Topics:
●Devices
●Build
●Entertainment
●Technology
●Open Source
●Science
●YRO
●Follow us:
●RSS
●Facebook
●LinkedIn
●Twitter
●
Youtube
●
Mastodon
●Bluesky
Become a fan of Slashdot on Facebook
Forgot your password?
Close
This discussion has been archived.
No new comments can be posted.
Load All Comments
Full
Abbreviated
Hidden
/Sea
Score:
5
4
3
2
1
0
-1
More
Login
Forgot your password?
Close
Close
Log In/Create an Account
●
All
●
Insightful
●
Informative
●
Interesting
●
Funny
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
bygreytree ( 7124971 ) writes:
... the prices of everything will come down when "Open"AI finally go bust in February ?
byallo ( 1728082 ) writes:
Would you bet on February?
bygreytree ( 7124971 ) writes:
Play it safe: March
https://youtu.be/oUDs7Oe8jyI?t=16
byallo ( 1728082 ) writes:
I'd think you can bet on a lot of companies both of us never heard of going bust in 2026 because they sell things with the only selling point "It's AI", but OpenAI will survive any crash. They are not just ChatGPT, but one of the largest infrastructure providers for LLM-APIs. GPT-5.2 Pro API price is Input $21/MToken and Output $168/MToken, that's surely making them more money than the consumer subscriptions.
bygreytree ( 7124971 ) writes:
"that's surely making them more money than the consumer subscriptions"
But it is not making more money than the ever-increasing amount they HAVE TO spend on building infrastructure, just to stay in the game.
Like waves that draw in water to grow, eventually there is no more water and the waves CRASH.
byallo ( 1728082 ) writes:
Some other letters in GAFAM weren't profitable in the first ten years either. The question is, if they make it until then.
The cost (model size, with that electricity and so in the end money) of LLM was going down drastically this year and this may continue for a while. The cost of model training of the large American companies is not known, but of course needs to be amortized. While the compute ratio is dominated by inference since some time this may change again if inference cost keeps going down and training cost may stay the same.
But in the long game the AI companies can only win. Hardware gets cheaper, models get better, but the price customers are willing to pay stays the same, in particular if the models get better, even though it may be a more efficient but better model. For APIs it is not that sure, because $20 is a flatrate price for ChatGPT that is acceptable for users who trust the brand, but with APIs there is competition about the price per token. On the other hand some model characteristics are a minor lock-in, which means that API customers will not jump between providers for a few cents. But long-term the companies will compete over the API prices.
Mostly for the customer interfaces, the companies also start to use more interface tricks and fewer model calls, which isn't as bad as it sounds. If the chatbot first accesses a Wikipedia dump, getting the information is cheaper, but at the same time it may have less hallucinations if it is grounded with a (more or less) trusted knowledge source. Google search has its knowledge graph for a decade, I think it is trivial for them to add it to an AI query just as they add it to a search query.
Google's wide range of services and ownership of (user) data is probably also why they bootstrapped Gemini that good after it first looked like they are failing to compete in the field. Do you remember their announcements about Bard and how it was called before? The first year they looked like Apple currently does and didn't seem to get started, now OpenAI talks about "code red" when there is a new Gemini version.
Parent
twitter
facebook
There may be more comments in this discussion. Without JavaScript enabled, you might want to turn on Classic Discussion System in your preferences instead.
Slashdot
●
●
Submit Story
If A = B and B = C, then A = C, except where void or prohibited by law.
-- Roy Santoro
●FAQ
●Story Archive
●Hall of Fame
●Advertising
●Terms
●Privacy Statement
●About
●Feedback
●Mobile View
●Blog
Do Not Sell or Share My Personal Information
Copyright © 2026 Slashdot Media. All Rights Reserved.
×
Close
Working...
Working...