AI costs how much? GitHub Copilot users react to new usage-based pricing system

God forbid people realize the cost of the thing they've been using. This was, of course, inevitable.

All that VC / hyperscaler money that had been getting burnt on customer acquisition and experimenting was going to demand ROI at some point.. and that point is now.

2026: The year the AI bill came due.

It turns out you can't run a massive cash burn hopes-and-dreams machine forever without consequences. It's no small coincidence that that OpenAI / Anthropic are rushing to IPO before the costs catch up with everyone waking up to the real costs of the LLMs.
 
Last edited:
Upvote
638 (640 / -2)

cadence

Ars Scholae Palatinae
1,006
Subscriptor++
Coder Henri Kinnunen writes that they only burned 161 credits in a “productive day” of using Claude 5.3-Codex through Copilot

There is no such thing as Claude 5.3-Codex. They were most likely talking about GPT. But even then, it doesn't make much sense to use GPT 5.3-codex these days if you want to save money on tokens.
 
Upvote
100 (102 / -2)

cadence

Ars Scholae Palatinae
1,006
Subscriptor++
The million (billion? trillion???) dollar question is how much does it actually cost to run these models. I think companies like Anthropic and OpenAI earn a big profit on customers who pay per token. I'm looking forward to their IPOs so that I can see how their finances really look like.
 
Upvote
60 (83 / -23)
Post content hidden for low score. Show…

Hypatia

Ars Centurion
270
Subscriptor
Will this lead to a cascade that finally “corrects” the speculative bubble?

I have no sympathy for the billionaire techbro crowd, but I do worry for the many people whose 401k accounts are likely wrapped up in one way or another with the private debt market that has been leveraged into the Ai bubble. When the bubble pops (and it needs to), many people who aren’t billionaire techbro types will see pain.

For that reason alone, screw the billionaires. Add to that the direct damage to communities that did not ask for and do not want data centers, and double screw the billionaires.
 
Upvote
293 (301 / -8)

Don Reba

Ars Praefectus
3,339
Subscriptor++
The million (billion? trillion???) dollar question is how much does it actually cost to run these models. I think companies like Anthropic and OpenAI earn a big profit on customers who pay per token. I'm looking forward to their IPOs so that I can see how their finances really look like.
I have a feeling they are still losing money, and even these new prices are just easing users into paying the real cost of AI.
 
Upvote
340 (342 / -2)

Sarty

Ars Tribunus Angusticlavius
7,978
This reminds me of the pricing scheme of SMS/text messaging in the nascent cell-phone days.
No, totally backwards! The actual content of a text message is tens of kilobytes, and it ought to be ~free relative to voice-quality audio. It was always ridiculous to get charged by the text--I suspect the tower-to-phone handshake traffic was higher bandwidth.

This is more like, "It's one question to the slopbot, Michael. What could it cost, ten dollars?" and the answer might be yep.
 
Last edited:
Upvote
191 (197 / -6)
My gut feeling: It's still being subsidized, even at that cost :O I expect the following to happen to "GenAI" :

1. The complex AI tools will be used for niche purposes that do not generate a large enough profit for the large companies like Softbank to get a return on investment. Only Microsoft, Google, and AWS will be left standing, and OpenAI, et.al. will be absorbed by them.

2. For simple LLMs, I expect users to move to free local tools using frontends like Jan or GPT4All, using models like OpenAI OSS-20B.

3. I expect Sam Altman et. al. to eventually be brought up on felony fraud charges. I believe they made rich people, like Softbank, Microsoft, and Google, angry with their belief that we are "thissss close" to AGI, and they will be out for blood.
 
Upvote
135 (144 / -9)
Post content hidden for low score. Show…
This reminds me of the pricing scheme of SMS/text messaging in the nascent cell-phone days.
The technology used for SMS/ text messaging is completely differently, and prices can be reduced. I can not see how pricing can be reduced for complex LLM usage. Can you explain how cost can be lowered using actual examples with sources?
 
Upvote
8 (26 / -18)

Fatesrider

Ars Legatus Legionis
25,411
Subscriptor
The million (billion? trillion???) dollar question is how much does it actually cost to run these models. I think companies like Anthropic and OpenAI earn a big profit on customers who pay per token. I'm looking forward to their IPOs so that I can see how their finances really look like.
It varies, even to the cost per token, since different companies have different methods of energy generation and costs associated with just keeping the lights on, people employed and plans moving forward.

The actual amount is likely far higher than they've been charging, since not one AI company has achieved profitability after five years of this bullshit.

But this is where the rubber meets the road when people are being charged the ACTUAL COSTS for their use. To date, no one has been.

Now that at least ONE AI's "I need to make SOME profit" pricing has been publicly revealed, the future of AI and the actual costs of it can be more properly evaluated.

This ONE case isn't necessarily typical, but it is indicative that they've been grossly low-balling the costs per token to the customer in order to try to hook them on the drug and get that addiction/reliance going before telling everyone to pay up.

I'd GUESS that they priced higher than they had to just to get some revenue coming in, knowing that a portion of the users are going to Nope! out of using it from now on, or cutting way the fuck back. Cutting back wouldn't NECESSARILY be bad for the AI company, since presumably, the full costs of the token plus profit are included in each one (like it should have been from the beginning). So even a lower use amount will return a profit - assuming the income is greater than the overall operating costs per month.

It's very likely that a lot of these places may not be able to keep customers at a minimum price point, because they have to charge enough to stay solvent. And increasing the cost per token to cover costs, even without a profit, would likely result in more people abandoning the AI ship.

Casual users are going to vanish, so they'll never see a dime from them. The rest have to have a reason for it, and THOSE costs can't be higher than it was to just have the employees doing what they did before without AI, or they'll just hire new people and dump the AI, or cut back to the point the AI can't stay in business anyhow. They can reduce the cost of a token to the customer, but they can't make the customer pay more than the market will bear, either.

This is where the fun begins Reality is inserting itself into the narrative. It's going to be interesting to see what happens next.
 
Upvote
62 (74 / -12)
No, totally backwards! The actual content of a text message is tens of kilobytes, and it ought to be ~free relative to voice-quality audio. It was always ridiculous to get charged by the text--I suspect the tower-to-phone handshake traffic was higher bandwidth.

This is more like, "It's one question to the slipbot, Michael. What could it cost, ten dollars?" and the answer might be yep.
I was going to say the same thing. If I recall correctly, the sms system (at least originally) piggy backed on top of already existing technology, and it cost telecoms practically nothing to implement it. Charging to use it was a pure cash grab.
 
Upvote
155 (155 / 0)

Danathar

Ars Praefectus
4,560
Subscriptor
The frontier models are interesting, but of late I've been paying close attention to the progress of the CHEAPER models. Thats way more interesting in my mind.

If you understand their limitations, and you know how to guide them, the cheaper models can do quite a lot at MUCH lower cost.

If a particular model is able to get the job done, I'm VERY happy when a new model comes out and the one I've been using gets cheaper.
 
Upvote
64 (71 / -7)
Workers will be replaced by AI, businesses will adopt it and it will become essential for their processes, and then the price will increase dramatically until it’s almost/just as expensive as it used to be when meatbags did the work. Except now all that money is being siphoned into a handful of megacorporations instead of going to countless millions of workers in the form of salaries.

This was always the plan. Like Uber burning VC cash to subsidise rides until the taxi industry died, then jacking up the prices to reach profitability.

The billionaires and future-trillionares are no longer content with selling you products and services. They’re now literally coming for your entire salary.
 
Last edited:
Upvote
204 (213 / -9)
Workers will be replaced by AI, businesses will adopt it and it will become essential for their processes, and then the price will increase dramatically until it’s almost/just as expensive as it used to be when meatbags did the work. Except now all that money is being siphoned into a handful of megacorporations instead of going to countless millions of workers in the form of salaries.

This was always the plan. Like Uber burning VC cash to subsidise rides until the taxi industry died, then jacking up the prices to reach profitability.
The difference between Uber and these "GenAI" companies is that, unless they can "lock in" the end user, the end user will just leave for another provider. Another problem is that LLMs do not scale from a cost perspective; the more complex the model, the more it cost to use. The last problem is that it's unknown how much money it costs for the service before the quarry is run, and unpredictability is the bane to a business.
 
Upvote
88 (90 / -2)
Will this lead to a cascade that finally “corrects” the speculative bubble?

I have no sympathy for the billionaire techbro crowd, but I do worry for the many people whose 401k accounts are likely wrapped up in one way or another with the private debt market that has been leveraged into the Ai bubble. When the bubble pops (and it needs to), many people who aren’t billionaire techbro types will see pain.
My fear is that this will result in a government bailout of AI companies.
 
Upvote
107 (110 / -3)
My fear is that this will result in a government bailout of AI companies.
It depends on who is in charge after November. If it's the Republicans, I can see a bailout happening. If it's the Democrats, no way. Silicone Valley put all their eggs in the Republican basket, and it's going to come back to bite them.
 
Upvote
7 (30 / -23)