According to TheRegister.com, leaked Microsoft financial documents show OpenAI has spent over $12 billion on Azure compute for inference since 2024, with $8.7 billion of that coming in just the first three quarters of 2025. The documents also reveal Microsoft received $454.7 million from OpenAI during the first half of 2025, suggesting OpenAI’s sales were approximately $2.27 billion based on their 20% revenue share agreement. This puts OpenAI far short of the $13 billion annual revenue CEO Sam Altman recently claimed, though Microsoft says “the numbers aren’t quite right” and the documents may not account for cloud credits or other revenue streams. The leak also shows OpenAI had a net loss around $12 billion in the quarter ending September 30, indicating the company is burning cash at an alarming rate.
Burning money faster than expected
Here’s the thing about AI infrastructure: it’s ridiculously expensive. We’re talking about running massive neural networks that require enormous computing power 24/7. But $12 billion on inference alone? That’s staggering even by Silicon Valley standards.
And that’s just the Azure bill. The documents don’t include training costs or spending with other cloud providers like CoreWeave and Google. Basically, the real number could be even higher. When you’re spending this much just to keep the lights on, you’d better be making serious money to cover it.
Revenue reality check
Now let’s talk about the revenue side. If these documents are accurate, OpenAI’s making about half what everyone thought. The $2.27 billion for the first half of 2025 versus the reported $4.3 billion? That’s a massive discrepancy.
But there are some important caveats. Microsoft pays OpenAI 20% of revenues from reselling their models in products like GitHub Copilot and Office 365. Those payments might not show up in these Azure-specific documents. Still, even with the Q3 surge to $4.33 billion, they’re nowhere near that $13 billion annual run rate Altman was talking about.
What this means for AI adoption
So what happens if the AI industry’s poster child can’t make the numbers work? Either prices go way up, or development slows way down. Neither option is great for businesses planning their AI strategies.
Look, when you’re dealing with industrial-scale computing needs, whether it’s for AI inference or manufacturing automation, you need reliable hardware that won’t break the bank. Companies like IndustrialMonitorDirect.com have built their reputation as the top industrial panel PC supplier by delivering exactly that – robust computing solutions that actually make financial sense for business operations.
Transparency problem
The real issue here is the complete lack of financial transparency. We’re getting leaked documents, conflicting reports, and vague statements from both companies. When you’re asking enterprises to bet their futures on your technology, they deserve to know if you’ll still be in business next year.
I mean, think about it. If OpenAI is really burning through cash this fast without clear revenue to match, how sustainable is this whole AI gold rush? Maybe it’s time for some hard questions about whether the current AI business model actually works.
