Post

Marcus Hutchins
@malwaretech.com

I was very bearish on AI to begin with, but recently I started experimenting more with APIs and building my own agents. And wow, I did not realize how heavily subsidized all the subscriptions were: 🧵 1/6

Marcus Hutchins
@malwaretech.com

The $200/month Claude subscription I have is okish. Current frontier model capabilities are mediocre at best, but I can find $200/month worth of value here and there, so it's worth it for me. But, via the API though, I can burn $200 worth of tokens with a single prompt. 2/6

3 reposts73 likes

Sean
@feeney.dev

You can turn on otel logging and push your claude code usage to Prometheus / Grafana to see your actual CC usage if you are really interested in tracking that usage. It's proved useful to demonstrate to my friends just how subsidized this current set is.

Marcus Hutchins
@malwaretech.com

Most of the die-hard AI fans I talk to are using stuff like Cursor, Claude Code, ChatGPT, etc. All of those are heavily-subsidized to the point where those $20/month and $200/month subscriptions they're paying for are letting them burn thousands of dollars worth of tokens. 3/6

Marcus Hutchins
@malwaretech.com

I'd guess some of the heavier subscription users are probably burning like $15-30k/month worth of resources. Then you have to consider the fact that the actual cost of API tokens may also be subsidized. These company are getting discount compute from cloud providers looking to expand. 4/6

Johannes Weiss
@jweiss.io

Do you have recommended reading here? My understanding is that the subs are super subsidised indeed. But the actual API calls (which is what many enterprise customers are paying) don't immediately feel very subsidised to me. Are you saying AWS/GOOG are providing Anthropic with subsidised capacity?

Jake Williams
@malwarejake.bsky.social

The API token cost is *absolutely* subsidized by investors doing loss lead bets. But the more you squint, the dumber it gets. In talking with one of the largest players earlier this year, they noted that it's nearly impossible to calculate the true cost of serving tokens once subsidies stop.

Marcus Hutchins
@malwaretech.com

Yeah, it's crazy. Every time I bring it up people are just like "well technological advancements will make it cheap". How? Like what is your basis for this? And then they just hand wave it away

Marcus Hutchins
@malwaretech.com

So, like, yeah, "everyone gets a free supercomputer" is great and all. But it isn't a serious business model. I just don't see how any amount of capability advancement or decrease in operational cost makes generative AI even remotely good value for money. 5/6

Marcus Hutchins
@malwaretech.com

Then every AI proponent I talk to about this just come up with some insane napkin math where token costs keep decreasing exponentially forever. It feels as if the entire industry is basically operating on some insane Moore's law hypothesis where everything just becomes free... 6/6

KiloQubit
@kiloqubit.com

Google did find a way to reduce memory usage by 6x. But the only way I see LLMs being affordable is local models that run on your own device.

alias (in chains)
@antiali.as

No doubt the API is more expensive but $200 is like 20 million tokens at Opus 4.7 rates. What are you trying to do with a single prompt that takes so much?