What is OpenAI?
I realize you might say "a foundation model lab" or "the company that runs ChatGPT," but that doesn't really give the full picture of everything it’s promised, or claimed, or leaked that it was or would be.
No, really, if you believe its leaks to the press...
- OpenAI is a social media company, this week launching Sora 2, a social feed entirely made up of generative video.
- OpenAI is a workplace productivity company, allegedly working on its own productivity suite to compete with Microsoft.
- OpenAI is a jobs portal, announcing in September it was "developing an AI-powered hiring platform," which it will launch 'by mid-2026.
- OpenAI is an ads company, and is apparently trying to hire an an ads chief, with the (alleged) intent to start showing ads in ChatGPT "by 2026."
- OpenAI is a company that would sell AI compute like Microsoft Azure or Amazon Web Services, or at least is considering being one, with CFO Sarah Friar telling Bloomberg in August that it is not "actively looking" at such an effort today but will "think about it as a business down the line, for sure."
- OpenAI is a fabless semiconductor design company, launching its own AI chips in, again, 2026 with Broadcom, but only for internal use.
- OpenAI is a consumer hardware company, preparing to launch a device by the end of 2026 or early 2027 and hiring a bunch of Apple people to work on it, as well as considering — again, it’s just leaking random stuff at this point to pump up its value — a smart speaker, a voice recorder and AR glasses.
- OpenAI is also working on its own browser, I guess.
To be clear, many of these are ideas that OpenAI has leaked specifically so the media can continue to pump up its valuation and continue to raise the money it needs — at least $1 Trillion over the next four or five years, and I don't believe the theoretical (or actual) costs of many of the things I've listed are included.
OpenAI wants you to believe it is everything, because in reality it’s a company bereft of strategy, focus or vision. The GPT-5 upgrade for ChatGPT was a dud — an industry-wide embarrassment for arguably the most-hyped product in AI history, one that (as I revealed a few months ago) costs more to operate than its predecessor, not because of any inherent capability upgrade, but how it actually processes the prompts its user provides — and now it's unclear what it is that this company does.
Does it make hardware? Software? Ads? Is it going to lease you GPUs to use for your own AI projects? Is it going to certify you as an AI expert? Notice how I've listed a whole bunch of stuff that isn't ChatGPT, which will, if you look at The Information's reporting of its projections, remain the vast majority of its revenue until 2027, at which point "agents" and "new products including free user monetization" will magically kick in.
OpenAI Is A Boring (and Bad) Business
In reality, OpenAI is an extremely boring (and bad!) software business. It makes the majority of its revenue selling subscriptions to ChatGPT, and apparently had 20 million paid subscribers (as of April) and 5 million business subscribers (as of August, though 500,000 of them are Cal State University seats paid at $2.50 a month).
It also loses incredibly large amounts of money.
OpenAI's Pathetic API Sales Have Effectively Turned It Into Any Other AI Startup
Yes, I realize that OpenAI also sells access to its API, but as you can see from the chart above, it is making a teeny tiny sliver of revenue from it in 2025, though I will also add that this chart has a little bit of green for "agent" revenue, which means it's very likely bullshit. Operator, OpenAI's so-called agent, is barely functional, and I have no idea how anyone would even begin to charge money for it outside of "please try my broken product."
In any case, API sales appear to be a very, very small part of OpenAI's revenue stream, and that heavily suggests a lack of interest in integrating its models at scale.
Worse still, this effectively turns OpenAI into an AI startup.
Think about it: if OpenAI can't make the majority of its money through "innovating" in the development of large language models (LLMs), then it’s just another company plugging LLMs into its software. While ChatGPT may be a very popular product, it is, by definition (and in its name!) a GPT wrapper, with the few differences being that OpenAI pays its own immediate costs, has the people necessary to continue improving its own models, and also continually makes promises to convince people it’s anything other than just another AI startup.
In fact, the only real difference is the amount of money backing it. Otherwise, OpenAI could be literally any foundation model company, and with a lack of real innovation within those models, it’s just another startup trying to find ways to monetize generative AI, an industry that only ever seems to lose money.
As a result, we should start evaluating OpenAI as just another AI startup, as its promises do not appear to mesh with any coherent strategy, other than "we need $1 trillion dollars." There does not seem to be much of a plan on a day-to-day basis, nor does there seem to be one about what OpenAI should be, other than that OpenAI will be a consumer hardware, consumer software, enterprise SaaS and data center operator, as well as running a social network.
As I've discussed many times, LLMs are inherently flawed due to their probabilistic nature."Hallucinations" — when a model authoritatively states something is true when it isn't (or takes an action that seems the most likely course of action, even if it isn't the right one) — are a "mathematically inevitable" according to OpenAI's own research feature of the technology, meaning that there is no fixing their most glaring, obvious problem, even with "perfect data."
I'd wager the reason OpenAI is so eager to build out so much capacity while leaking so many diverse business lines is an attempt to get away from a dark truth: that when you peel away the hype, ChatGPT is a wrapper, every product it makes is a wrapper, and OpenAI is pretty fucking terrible at making products.
Today I'm going to walk you through a fairly unique position: that OpenAI is just another boring AI startup lacking any meaningful product roadmap or strategy, using the press as a tool to pump its bags while very rarely delivering on what it’s promised. It is a company with massive amounts of cash, industrial backing, and brand recognition, and otherwise is, much like its customers, desperately trying to work out how to make money selling products built on top of Large Language Models.
OpenAI lives and dies on its mythology as the center of innovation in the world of AI, yet reality is so much more mediocre. Its revenue growth is slowing, its products are commoditized, its models are hardly state-of-the-art, the overall generative AI industry has lost its sheen, and its killer app is a mythology that has converted a handful of very rich people and very few others.
OpenAI spent, according to The Information, 150% ($6.7 billion in costs) of its H1 2025 revenue ($4.3 billion) on research and development, producing the deeply-underwhelming GPT-5 and Sora 2, an app that I estimate costs it upwards of $5 for each video generation, based on Azure's published rates for the first Sora model, though it's my belief that these rates are unprofitable, all so that it can gain a few more users.
To be clear, R&D is good, and useful, and in my experience, the companies that spend deeply on this tend to be the ones that do well. The reason why Huawei has managed to outpace its American rivals in several key areas — like automotive technology and telecommunications — is because it spends around a quarter of its revenue on developing new technologies and entering new markets, rather than stock buybacks and dividends.
The difference is that said R&D spending is both sustainable and useful, and has led to Huawei becoming much a stronger business, even as it languishes on a Treasury Department entity list that effectively cuts it off from US-made or US-origin parts or IP. Considering that OpenAI’s R&D spending was 38.28% of its cash-on-hand by the end of the period (totalling $17.5bn, which we’ll get to later), and what we’ve seen as a result, it’s hard to describe it as either sustainable or useful.
OpenAI isn't innovative, it’s exploitative, a giant multi-billion dollar grift attempting to hide how deeply unexciting it is, and how nonsensical it is to continue backing it. Sam Altman is an excellent operator, capable of spreading his mediocre, half-baked mantras about how 2025 was the year AI got smarter than us, or how we'll be building 1GW data centers each week (something that, by my estimations, takes 2.5 years), taking advantage of how many people in the media, markets and global governments don't know a fucking thing about anything.
OpenAI is also getting desperate.
Beneath the surface of the media hype and trillion-dollar promises is a company struggling to maintain relevance, its entire existence built on top of hype and mythology.
And at this rate, I believe it’s going to miss its 2025 revenue projections, all while burning billions more than anyone has anticipated.