Interesting claim from SemiAnalysis.

AI subscriptions are dramatically underpriced versus API usage:

- For heavy coding/chat users, the subscription can be 40–70× cheaper than paying API rates; the API is mainly better when you need automation or product integration.

- a $200/month ChatGPT Pro plan can provide about $14,000/month of API-equivalent usage, while a $200/month Claude Max 20x plan can provide about $8,000/month.

this is about what i thought. people regularly used to report $10k+ API equivalent via ccusage

obviously, maximally utilized subscriptions are negative. but I think the negative margin isn't that dramatic anymore (because it's too easy to max out with agents now). I'd say current subscriptions must be profitable up to ≈35% utilization.

@teortaxesTex What if it’s part of their morality? Subsidize and support the little guy. No permanent underclass for the poor who can’t pay that much

The numbers are ~directionally right, but in my own testing (and GPT 5.5's own estimations), the OpenAI Pro 20x sub currently gives you precisely $2600 weekly GPT-5.5 API equivalent. Which is still huge. When I had the 5x sub (when it was still 10x), I also got precisely $1300/week, so right now it's about $650/week. And 5-hour limit for the 20x sub right now seems to be about $390-$410.

@RandomCSFan1 Lol they're not subsidizing, this is their main revenue stream

@teortaxesTex I've been saying this forever. No way it costs what they say it costs. I think they overpriced it at the API level to pretend subs were subsidized to create a narrative that encouraged subs.

@rohanpaul_ai Yea and it’s going away quickly. I easily use mid 5 figures of tokens on my max plan

@rohanpaul_ai I immediately started imagining the code to get responses from the website instead of the API, while doing most of the things the API can do.

I imagine it's a combination of subsidization and preventing affordable distillation.

vendor lock-in strats so that when it comes time to decide on seats they choose X company. Also the end goal has always been api billing. It's why anthropic is giving us limited time access to this "mythos" class model. Fast models were the first taste this is the first time they've started locking a class of models behind api billing. The days of the general public on the bleeding edge are coming to an end just in time for IPO.

@rohanpaul_ai $200 turning into thousands of dollars of effective usage is wild. No wonder the limits keep changing.

@rohanpaul_ai The pricing disparity here just underlines how much 'automation' is still human-in-the-loop copy-paste, not true agent integration.

@rohanpaul_ai Pricing gap this wide means the real battleground is margin compression on API compute, not subscription revenue.

@andersonbcdefg $200 soft cap was a myth the whole time

people were really maxing those cards in secret