Ars Technica reported leaked 2025 OpenAI financial documents showing roughly $13 billion in revenue, about $7.5 billion in cost of revenue, very large R&D spending, and nearly $6 billion in sales and marketing. The immediate reaction was not shock at the losses. It was that subsidized AI usage, huge training bills, and aggressive go-to-market spending already made this outcome obvious. Where people landed was sharper than that. The useful split was not "profitable or doomed". It was whether the numbers imply a business with decent unit economics buried under a temporary land grab, or a treadmill where the spending never really comes down.
A lot of people read the leak as better than expected. If the gross revenue and cost-of-revenue figures are even directionally right, OpenAI appears to have a real business, not a science project with no demand. Several comments framed this as standard hypergrowth behavior. You lose money while buying compute, researchers, and distribution in a market where the winner could capture a massive share of enterprise spend. Some also argued that comparing subscription sticker prices to
API rates misses how businesses buy value. If a better model unlocks work that a cheaper model cannot do, or gets there with fewer tokens, the economics can still work even at much higher nominal prices.
The harder view was that AI labs do not get to cleanly separate "
inference profits" from "R&D costs" the way bulls want to. Their product decays unless they keep training frontier models, improving inference efficiency, and shipping better tools. In that framing, R&D is not discretionary overhead. It is the cost of staying sellable. That point got tied to weak lock-in, fast model substitution, and pressure from cheaper open or Chinese models. A recurring claim was that customers will not pay frontier premiums forever when "good enough" keeps getting cheaper, especially outside high-value enterprise use cases.
The other major thread was monetization. The reported sales and marketing bill looked enormous, but commenters gave a mundane explanation. Consumer AI is not differentiated enough to market itself, and enterprise AI absolutely does require large sales teams, solutions engineers, and partner channels. Several people also expected ad-supported AI to expand because free users are plentiful and hard to convert. That idea came with skepticism. Search-style ad economics rely on concentrated distribution and user intent, while conversational systems have weaker moats and much trickier incentives once ads are blended into answers.
The mood was skeptical of the hype but not dismissive of the category. Most people seemed to accept that OpenAI can keep growing revenue and that frontier models genuinely create value for some users. The disagreement was over durability. The optimistic case says these losses are normal for a company trying to lock up a giant market under compute scarcity. The bearish case says the scarcity is not a moat if competitors catch up, pricing compresses, and the company must keep spending at full tilt just to stand still.