AI business models about to fail? Tech blogger digs into OpenAI’s “financial black hole”: money is burned three times faster than public data suggests, revenue is exaggerated and cannot cover costs!

```

A document purportedly from within OpenAI is posing a grave challenge to the financial status of this AI giant and to the business model of the entire generative AI industry.

The contents of the document show that OpenAI's actual operating costs may far exceed what the public imagines, while its revenue has been significantly exaggerated, creating a startling gap between its high expenses and income.

The source of this upheaval is well-known tech blogger and author of the newsletter "Where's Your Ed At," as well as a sharp critic of the tech industry—Ed Zitron. On one hand, he uses personal media platforms to fiercely criticize the AI industry, predicting its bubble is about to burst; on the other hand, his PR company EZPR has provided services to many AI startups. This dual role of both criticizing AI and representing AI companies makes him a controversial voice in the tech world. As an "insider" who understands the industry's inner workings, he firmly believes that the current AI boom is essentially a capital-driven bubble, and with his characteristic rhetoric, he predicts the industry’s "day of reckoning" is coming.

Zitron published an article on his blog, citing document data he claims to have seen stating that just OpenAI's model inference costs on Microsoft's Azure cloud platform have reached staggering levels. For example, in the first half of 2025, this expense was close to $5 billion, whereas media reports previously cited "cost of revenue" in the same period as only $2.5 billion. This suggests that OpenAI's cash burn rate could be nearly three times higher than public data shows.

Meanwhile, these files also reveal a 20% revenue share that OpenAI pays to its main investor, Microsoft. By reverse-engineering this data, OpenAI's actual revenue is far lower than previously reported numbers. For example, Microsoft's share for 1H 2025 implies OpenAI revenue of about $2.27 billion, vastly less than media reports stating $4.3 billion. If these numbers are true, not only is OpenAI's financial health worrying, but the entire AI large model industry's profitability will be thrown into serious doubt.

When it comes to the accuracy of this data, Microsoft and OpenAI have been vague in their responses. It is understood that a female Microsoft spokesperson told the Financial Times that "the numbers are not entirely accurate" but refused to provide more details. OpenAI declined to comment, merely suggesting the media verify with Microsoft. A person familiar with OpenAI said the numbers failed to provide "a complete picture." While no official confirmation can be obtained, the parties involved have not provided strong evidence to substantively refute the data.

Shocking Cost Black Hole: Inference Expenses Far Exceed Revenue

According to data released by Ed Zitron, there is a huge gap between OpenAI's operating costs (specifically model inference costs) and its revenue, and this gap is greater than any previously reported. Inference refers to the process by which applications like ChatGPT call large language models to generate responses, and it is one of its core operating costs.

The data shows that over the seven quarters from Q1 2024 to Q3 2025, OpenAI’s inference computation spending on Azure alone exceeded $12.4 billion. In just the first nine months of 2025, its inference cost reached $8.67 billion. In comparison, The Information previously reported that OpenAI's inference cost for all of 2024 was about $2 billion, with "cost of revenue" in the first half of 2025 at $2.5 billion. Zitron’s data shows the real cost is almost three times these reported figures.

More worrying is that these astronomical costs seem not to be in sync with revenue growth. The data shows that even factoring in revenue growth, inference spending seems to be rising linearly at a faster rate, completely overwhelming income. This pattern raises questions as to whether, with current technology and pricing, large model business can ever be profitable.

Exaggerated Revenue? Huge Discrepancies with Publicly Reported Income Data

Aside from high costs, the documents revealed by Zitron cast serious doubt on OpenAI’s revenue. Through Microsoft’s 20% revenue share, it is possible to reverse-calculate OpenAI’s minimum revenue, which is much lower than numbers cited by the media or disclosed by OpenAI itself.

2024: The document shows Microsoft received $493.8 million in revenue share that year, implying OpenAI's revenue was at least $2.469 billion. However, CNBC and The Information reported that OpenAI’s revenue in 2024 was expected to reach between $3.7 billion and $4.0 billion.

First half of 2025: Microsoft received $454.7 million in revenue share, suggesting OpenAI's revenue in that period was at least $2.273 billion. However, The Information reported OpenAI generated $4.3 billion in revenue in that timeframe.

These huge differences are hard to reconcile. Even OpenAI CEO Sam Altman’s claim that the company’s annualized income "well exceeds" $13 billion is at odds with the financial situation revealed by the documents. Zitron speculates the high revenue numbers reported in the media may result from creative calculations of "annualized revenue" or "annual recurring revenue" (ARR), but OpenAI has never clarified its definition.

Complex Partnerships and Vague Responses

It is noteworthy that OpenAI’s financial ties with Microsoft are extremely complicated, not a simple investment and cost relationship. Reportedly, there is a two-way revenue sharing agreement between the two. In addition to taking 20% of OpenAI’s revenues, Microsoft must also pay OpenAI 20% of income earned through Azure OpenAI service model sales. There are also Bing-related revenue shares and possible royalty payments.

This complex structure means that calculating OpenAI's total revenue based solely on the revenue share received from Microsoft will inevitably lead to an "undervaluation". Yet even considering this, the huge gap between cost and revenue revealed by Zitron's data remains hard to explain.

It is understood that when approached for confirmation, the Financial Times did not receive clear denials from either company. Microsoft's statement that "the numbers are not entirely accurate" and a source’s claim that "the full picture is not available" leave much room for market speculation (or concern). Without more specific rebuttals, the credibility of these leaked figures is growing.

Industry Alarm: The Sustainability of the AI Business Model is Being Challenged

Even if Zitron's disclosed data is only partially accurate, it sounds the alarm for the entire generative AI industry. This suggests that the business model of industry leader OpenAI may simply not be sustainable.

Based on these numbers, The Financial Times’ analysis shows that at current growth rates, OpenAI's minimum estimated revenue may only cover its inference costs around 2033. If Microsoft’s 20% revenue share is deducted, it might "never" be able to cover inference costs with its own income.

This finding raises a sharp question for investors and the market: if even the most well-funded and leading company, OpenAI, faces such tremendous financial pressure, what will happen to other general large model providers? Analysts believe there are only two possible futures: either model operating costs must plummet, or customer-facing pricing must rise significantly. However, so far there are no signs of either trend, casting a heavy shadow over the commercialization prospects of generative AI.

Risk DisclaimerMarkets have risks, investment needs caution. This article does not constitute personal investment advice, nor does it take into account individual users' unique investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article suit their particular circumstances. Investing based on this information is at your own risk. ```