OpenAI's Insane Scaling Problem
$20 billion?
A few days ago, OpenAI announced that its Annualised Recurring Revenue (ARR) for 2025 was $20 billion. That is more than double their 2024 annual revenue! So, has Altman finally succeeded? Is OpenAI close to becoming a financially stable company? Well, no. For one, this $20 billion figure seems awfully optimistic when you look at the details. But even if they did actually achieve this figure, it likely isn’t enough. Why? Because AI has a serious scaling problem. Let me explain.
Things Aren’t Adding Up.
In the first half of 2025, OpenAI reportedly generated $4.3 billion in revenue, which was expected to reach $12 billion by the end of the year. So, with an annual revenue of $20 billion, they generated twice as much money in the second half of 2025 as they had anticipated. What drove this dramatic increase? Weekly users and paid users haven’t increased substantially more in the second half of 2025 than in the first, and there have been no new major corporate partnerships to drive this level of revenue either. What is going on?
Well, as found by Ed Zitron, a telltale sign that things are not what they seem is that OpenAI paid Microsoft $454.7 million in revenue sharing during the first half of 2025. As part of their partnership, OpenAI is expected to pay Microsoft 20% of its revenue. So that means OpenAI had actually made $2.27 billion in revenue during the first half of 2025, which is $2 billion less than reported. In Q3, OpenAI paid Microsoft another $411.1 million, meaning they made $2 billion in revenue in Q3. All the figures suggest that Q4 could have experienced a similar growth, so we can assume a $4 billion revenue for Q4. But all of this equates to an annual revenue of around $9 billion, which is less than half the claimed $20 billion.
Another sign that things aren’t quite what they seem is OpenAI’s reported users. OpenAI has reported that ChatGPT had 800 million weekly users by the end of 2025. Multiple third-party sources have found that only 5% of users pay for ChatGPT, so we know that roughly 40 million people do. If they all pay the $20 per month subscription, that equates to $9.6 billion a year in revenue. But we also know that roughly 30% of OpenAI’s revenue comes from other sources, like licensing. So we can infer that the current annualised revenue should be around $13.7 billion. However, ChatGPT started 2025 with far fewer weekly users, which means that its total revenue for 2025 should be substantially below $13.7 billion.
This is why I personally find that $20 billion figure hard to believe. It looks like $10 billion or more has magically dropped into OpenAI’s wallet at some point in the last quarter of 2025. Where did that come from?
The Microslop Loophole?
Needless to say, the only thing this proves is that OpenAI needs to be far more financially transparent. There is not enough information here to build a clear picture of what is really going on. But because the publicly available data does not align with OpenAI’s narrative at all, it, at the very least, suggests something is afoot. I have only been able to establish one viable explanation for what might be going on here, but I must stress that this is my opinion of what might be causing this discrepancy.
Remember that 20% revenue share between OpenAI and Microsoft? Well, it is a two-way street. Microsoft pays OpenAI 20% of the revenue it generates from the sale of repackaged versions of OpenAI’s models, such as Copilot and Azure AI.
But Microsoft is really struggling to sell these AIs. For example, we know that their Azure AI based on ChatGPT has sold so poorly that they have slashed sales quotas in half. Likewise, rather than selling Copilot exclusively as an additional subscription, Microsoft has embedded it into existing subscriptions, such as Microsoft 365 (formerly Office). In fact, as a new customer in the UK, I cannot buy a 365 subscription without Copilot. Likewise, Microsoft has embedded ChatGPT into their Bing search engine.
Microsoft is really bad at financial transparency when it comes to how it makes money from selling these AI models. But it is most likely making very little from them. Their 2025 revenue was only 15% higher than 2024. And the vast majority of that increase came from their Azure cloud computing business, not from selling the products embedded with AI.
Integrating Copilot into essential subscriptions and services achieves two things. Firstly, it forces this terrible technology onto us, the public, which makes their numbers look better. And secondly, it means that Microsoft can technically claim that AI drives all the revenue generated from these subscriptions and services. As such, they couldsend 20% of it to OpenAI, even though their models didn’t directly create this revenue.
I suspect this is where that $10 billion-plus figure appeared from. OpenAI desperately needed cash, and Microsoft used this method to send it over covertly. To be clear, I cannot prove this, and I am not making an accusation. I just believe this might have happened.
But why would Microsoft do this? Again, it’s all down to OpenAI’s scaling problem.
More Money, More Problems.
Even if OpenAI genuinely made $20 billion in revenue in 2025, it wouldn’t be anywhere near enough. Multiple analyses have found that OpenAI’s operational costs will be significantly more than $28 billion in 2025. So, even with $20 billion in revenue, they are still likely miles away from breaking even, let alone creating sustainable profit.
Not to mention the fact that OpenAI’s costs could be even higher than previously imagined. You see, they recently announced they had 1.9 GW of computing power in 2025.
According to IBM CEO Arvind Krishna, a single GW of AI-capable data centres costs $80 billion to build. But we also know that a GW of AI data centres consumes around $1.3 billion in energy costs annually and that these data centres have a realistic operational lifespan of three years. Logically, the annual cost of a single GW of AI compute power (including the annually spread build cost and energy cost) is $27.97 billion. So OpenAI’s 1.9 GW of compute will cost around $53 billion annually.
We don’t know how much of this cost OpenAI shoulders. Some will be soaked up by the owners and operators of these data centres, which is why many AI data centre operators are running at a catastrophic loss. We also know that the amortisation of these data centres is likely being significantly underplayed, given that operators are claiming a five-year lifespan to make the books look better. So, how much of this loss is actually being identified is also questionable. Additionally, this 1.9 GW figure is likely referring to OpenAI’s total compute capacity at the end of 2025, meaning the total cost during 2025 was slightly lower.
Therefore, in the best-case scenario, where this $20 billion figure is honest revenue and OpenAI’s operational costs were as predicted, they would make an $8 billion loss in 2025. That is noticeably larger than the $5 billion loss they posted in 2024. But in the worst-case scenario, where their actual revenue is closer to $10 billion and their compute costs are accurately disclosed, they might face losses in the multiple tens of billions of dollars.
Quite simply, the costs are scaling quicker than the revenue.
Another Bailout?
Regarding that $5 billion loss in 2024, it was actually large enough to threaten OpenAI with bankruptcy. In fact, the only reason OpenAI survived to see 2025 was due to a $6 billion corporate bailout from its backers, mainly Microsoft. Microsoft had sunk tens of billions of dollars into OpenAI and had already begun basing much of its new direction on its partnership with OpenAI. In other words, if OpenAI went under, it would be disastrous for Microsoft. Bailing them out was likely the cheaper option, even if it damaged OpenAI’s reputation.
I think this might explain why it seems like more than $10 billion magically appeared in OpenAI’s wallet in Q4 2025. The figures suggest that OpenAI was on course to take a catastrophic loss by the end of the year that would be large enough not just to bankrupt them but also to pop the AI bubble. They would need another bailout to, at least, soften the blow. So, I suspect that in Q4 2025, Microsoft allocated a sizeable portion of its subscription income as AI-driven so that they could inconspicuously throw a huge lump sum in the direction of OpenAI under the guise of revenue sharing (despite AI-driven revenues being basically non-existent) to dramatically reduce this loss, which would protect OpenAI and Microsoft. Again, I cannot prove this, and this is not an accusation; I am merely saying that, in my opinion, I can see this as a possibility.
Needless to say, if this is proven correct down the line, it would be devastating.
The Problem
Economies of scale are a very important part of any business. This concept refers to how the unit cost of a good or service reduces as production scale increases. This reduction is driven by fixed overheads being further spread out, supply chains being optimised, and products being optimised. The only reason the average person can afford microwaves, phones, computers, EVs or similar is because of economies of scale.
But LLM AIs do not follow this trend. Scale does not reduce the unit cost. Furthermore, the cost of developing these AIs is increasing exponentially as they attempt to make them more capable (read more here). So, trying to give AI a larger scale (i.e., making it useful in more areas) can dramatically increase the unit cost.
In other words, the larger and better you try to make AI, the further away from profitability it becomes, given that costs scale up faster than revenue.
OpenAI is proving this rather beautifully. Even if they have genuinely more than doubled their annual income from 2024, which I highly doubt, their annual loss has grown by at least 33%. They are going backwards, even further into the red. You can only do that for so long before the lights are turned off.
Thanks for reading! Everything expressed in this article is my opinion, and should not be taken as financial advice or accusations. Don’t forget to check out my YouTubechannel for more from me, or Subscribe. Oh, and don’t forget to hit the share button below to get the word out!
Sources: CTech, The Register, The Register, AI Invest, The Register, Futurism, Bezinga, Investing.com, Macrotrends, TechCrunch, Futurism, Ed Zitron, Ed Zitron
Annualized recurring revenue means that in December they recorded $1.6B not that they hit $20 for the whole year.