OpenAI's full-year 2025 audited results paint a picture of rapid revenue growth that still falls far short of covering its ballooning expenses, leaving a substantial operating loss even before non-operating adjustments tied to its shift toward for-profit status.
apparently ed proved inference is VERY profitable, and the companies, in fact, do not lose money on subscriptions.
Exclusive: I have seen OpenAI's audited financials for 2024 and 2025.
In 2025, OpenAI had $13.07 billion in revenue and $34 billion in costs. $867 million of its revenue came from SoftBank, and $303 million came from Microsoft. https://www.wheresyoured.at/exclusive-openai-financials/
Nearly half the $34 billion in total costs went to research and development, including $10.59 billion paid to Microsoft for infrastructure, while sales and marketing added another $5.73 billion.
The figures surface just as OpenAI holds roughly $25 billion in cash and equivalents, yet no verified schedule or valuation details have been released to connect these 2025 results directly to a planned public offering.
Many users reacted negatively to OpenAI's reported $13B revenue against $34B costs, dismissing the business model as an unsustainable land grab with a deadline ahead of its planned IPO.
you can find estimates on what fraction of revenue OAI has been receiving from subs as compared to API, this has been reported and is based on leaked docs/financials.
GPT-5.5 pro estimates ~75% of revenue came from subs, based on reports from FT and The Information. Excluding enterprise seats, it predicts 65%.
65% * 13.07=$8.5B, which, as you can check via Google or your favourite LLM, is bigger than $7.5B
FT: OpenAI is growing fast, but its numbers show how expensive frontier AI has become.
They are spending $34B last year to buy compute, talent, and distribution.
The company booked about $13B in revenue, but spent about $19B on R&D and nearly $6B on sales and marketing.
---
ft .com/content/e15b0d7e-ff6b-4f16-ba7a-4068feddb828?syn-25a6b1a6=1
"stripping out stock based comp and the value of compute credits" why on earth would you do that if you wanted to learn if OpenAI has a profitable business model?
For now at least, they do in fact need to pay their staff, and compute is not free.
Not to say Zitron isn't a charlatan (he is) but that's equally silly.
@mooncat_is this is his core belief that he publicly held onto until halfway through '25. he just switched his line to financials after this became publicly untenable, while keeping his conclusion the same
@stalkermustang Да они at cost сервят в среднем подписки и на апи накрутка огромная. Но вот с другой стороны, а в какой степени стоимость тренинга и масштаб инференса функция от экономической полезности? Точно точно перевесит прибыль от инференса стоимость r&d?
@mooncat_is i think his point is that exponential growth of losses is a bad business and that altman lied
@stalkermustang then they might still lose on subs and profit from api
@stalkermustang If you don't pay your employees, your rent and you receive the model training free, then you don't lose money on inference!
https://chatgpt.com/s/t_6a312526d0308191b6189d2ff5319f3b
pro ext estimates for those who are curious, with 3 scenarios
The lower bound is 7b for all subs, which is lower than 7.5b, but the inference for all of API + free users is more than $0.5b/year (otherwise Ed & co would have to agree that the markup is like 1000% or thereabouts)
@thewhiteboxAI > (otherwise Ed & co would have to agree that the markup is like 1000% or thereabouts)
Which... would again kill the whole argument of unsustainable business, because API is soaring (subs are shrinking), meaning (rev-compute) is veeeery positive
@dh44t or you can start reading what's written. Ur choice
@stalkermustang @dh44t Even if you deduct the R&D budget they're still losing money. And none of these labs can make money for long unless they keep dumping that way.
You can't just remove Admin and Marketing budgets.
@CharlesD353 @ShakeelHashim Is stock-based comp expected to continue indefinitely post-IPO?
@GrigoryEvko может нет, может да - зависит от роста спроса. Но это не то, что писал Ed больше года :)
@sam90860759 that's not what's written in the post, you cunt. But I'm not surprised you can't read.
@mooncat_is thats not the point though
@Gusarich theoretically yes. IIRC, customers' revenue was more than half of the total, so it was bigger than API (esp. in 2025).
so it could be 7b revenue was made by, say, 7b cost of compute (at extreme).
Still not as a dramatc loss as ed depicts it
@stalkermustang @dh44t Personal attacks already, wow, very mature.
You also haven't proven the revenue is entirely inference. Licensing, consulting, any other business is also in that figure, and the surplus should definitely be calculated on lifetime as they can only serve it for ~2-3 years.
@stalkermustang Or ya let’s just ignore everything else bc they’re one time cost right? Fucking retards
But sir, you don't have to take Ed's numbers for granted, just look at what the Labs tell you.
Anthropic disclosed they were spending $13 per dev per day, averaging $150-$250/month. Most subscriptions are $20/month, so they are deep underwater, and even the max-20x subscriptions have, at best, a 25% gross margin. That is terrible, but even then, this is only the total picture; you have to add training costs (which Labs may or may not include in COGS), and ALL operating expenses, which include depreciation, salaries (including stock-based compensation, which is huge for Labs), and marketing and sales. Once you factor all of that in, the cost/token is super high, in the single-digit dollars per million tokens, which makes subscriptions unsustainable. I'm sorry, but a technology with high marginal acquisition costs like AI can't make money on subscriptions, ever.
The silver lining is that APIs are a much better business than most believe, but I genuinely don't care because with AI Labs, it's all about free cash flow. I don't care you're making $8 for every $10 you charge a customer if you're then spending $20.
