OpenAI's Leaked Financial Documents Paint a Complicated Picture
OpenAI, the artificial intelligence company behind ChatGPT and the GPT series of language models, has long been the subject of intense financial speculation. Now, thanks to leaked audited financial statements obtained by independent journalist Ed Zitron and reviewed by the Financial Times, the public has its clearest view yet of what is happening inside the company's books. The short version: revenue is growing at a breathtaking pace, but expenses are growing even faster — and the gap between the two is measured in billions of dollars.
With the company reportedly filing SEC paperwork ahead of an anticipated initial public offering, the timing of this disclosure couldn't be more significant. Investors, analysts, and technology watchers are now asking a fundamental question: can OpenAI ever turn its staggering costs into sustainable profit?
The Revenue Numbers Are Genuinely Impressive
Let's start with the good news for OpenAI. The company's revenue growth over the past two years has been extraordinary by almost any measure. According to the leaked documents, OpenAI's reported revenue climbed from $3.7 billion in 2024 to $13.07 billion in 2025 — representing a year-over-year increase of more than 250 percent. That is the kind of hypergrowth that venture capitalists dream about and that makes Wall Street sit up and pay attention.
The Financial Times added further context by noting that OpenAI's monthly revenues had grown to nearly $2 billion by the end of 2025, suggesting that the revenue trajectory continued to steepen throughout the year rather than plateauing. If that monthly run rate holds or accelerates into 2026, the company could be on track to generate revenues exceeding $24 billion on an annualized basis — a figure that would place it firmly among the top-tier technology companies in the world.
This growth is largely driven by the continued expansion of ChatGPT's subscriber base, the increasing adoption of the OpenAI API by enterprise customers, and the rollout of new premium products and services. The company has successfully monetized what was once a free research tool into a commercial platform with significant and growing demand.
The Expense Problem: When R&D Costs More Than You Earn
Here is where the financial picture becomes far more complicated. Despite the impressive revenue figures, OpenAI's expenses dwarf its income by a wide margin. The company's research and development costs alone — before even accounting for other operational expenses — exceeded total revenues in both 2024 and 2025.
In 2024, OpenAI reported R&D expenses of $7.81 billion against total revenues of $3.7 billion. In 2025, those R&D expenses ballooned to a staggering $19.18 billion, more than $6 billion higher than the company's $13.07 billion in revenues. To put that plainly: OpenAI is spending roughly $1.47 on research and development for every single dollar it brings in.
A significant portion of those R&D costs flowed directly to Microsoft. According to the leaked documents, OpenAI paid $10.59 billion to Microsoft in 2025 alone for R&D-related services. This figure reflects the deep and complex relationship between the two companies, with Microsoft providing the massive cloud computing infrastructure — primarily through its Azure platform — that OpenAI requires to train and run its increasingly powerful models. Microsoft has invested approximately $13 billion in OpenAI over the years, and clearly a substantial portion of that investment effectively cycles back to Microsoft in the form of infrastructure payments.
Why Does AI Training Cost So Much?
To understand why OpenAI's R&D costs are so enormous, it helps to understand what those costs actually represent. Training a frontier AI model like GPT-4 or its successors requires running tens of thousands of specialized graphics processing units — most commonly Nvidia's H100 and A100 chips — continuously for weeks or months at a time. The electricity costs alone for a major training run can reach into the hundreds of millions of dollars. Add to that the cost of leasing or purchasing the hardware, the salaries of world-class AI researchers, the data acquisition and licensing costs, and the ongoing costs of running inference at scale for millions of daily users, and the numbers begin to make more sense.
OpenAI is not simply maintaining existing products; it is engaged in an aggressive race to build and deploy the most capable AI systems in the world. Each new model generation tends to require significantly more compute than the last, which means expenses are likely to remain elevated — or grow further — as long as the company remains at the frontier of AI development.
What This Means for OpenAI's Expected IPO
The leaked financials arrive at a particularly sensitive moment. OpenAI is widely reported to be preparing for a public stock offering, which would represent one of the most anticipated IPOs in recent tech history. For that offering to succeed, the company will need to convince public market investors that its current losses are a temporary feature of a high-growth, high-investment phase rather than a structural problem with its underlying business model.
That will not be an easy case to make. Unlike earlier tech IPOs where companies were losing money on customer acquisition but had clear unit economics, OpenAI's losses are tied directly to the core activity that generates its revenue. The more models it trains, the more it costs — and the competitive pressure to keep training new models shows no signs of slowing.
The Bigger Question: Can AI Scale Into Profitability?
OpenAI's financial situation is in many ways a microcosm of the broader challenge facing the artificial intelligence industry. Enormous capital expenditure is required upfront, with profitability depending on achieving sufficient scale and on the eventual reduction in the cost of compute over time. Proponents of this view point to historical precedents in cloud computing and semiconductor manufacturing, where costs fell dramatically as the technology matured.
Skeptics, however, argue that the compute requirements for frontier AI are growing faster than the cost reductions, creating a treadmill effect where achieving profitability always seems just around the corner. The leaked OpenAI documents do nothing to definitively settle that debate, but they do confirm that the company is currently deep in the red — and that the path to profitability, while plausible, remains a work in progress.
For now, all eyes will be on OpenAI's SEC filings, its IPO timeline, and whether the company's monthly revenue trajectory can eventually close the enormous gap between what it earns and what it spends.