It was a quiet Tuesday morning in Seattle when a single spreadsheet, leaked from the heart of OpenAI’s finance team, began its viral journey across the internet. No dramatic hack, no shadowy whistleblower—just a file, quietly copied, quietly shared, quietly exposing the raw, beating heart of the AI revolution. What it revealed wasn’t just numbers. It was a story of ambition, cost, and a partnership that’s quietly reshaping the future of technology.
The Spreadsheet That Shook Silicon Valley
At first glance, the document looked like any other corporate report: rows of figures, dates, and vendor names. But as tech journalists and analysts dug deeper, a startling truth emerged. Microsoft, OpenAI’s biggest investor and cloud partner, had collected nearly $866 million from OpenAI by September 2025—just for letting the AI giant run its models on Azure, Microsoft’s cloud platform. That’s 20% of OpenAI’s revenue, paid straight to Microsoft, year after year.
But here’s the twist: OpenAI’s own costs for running its models—what experts call “inference”—had skyrocketed. In 2024, the company spent $3.8 billion just to keep its AI models answering questions. By the first nine months of 2025, that number had ballooned to $8.65 billion. For context, that’s more than the GDP of some small countries.
The Hidden Cost of Talking to AI
Imagine you’re a teacher in Chicago, using ChatGPT to help your students with homework. Every time you type a question, OpenAI’s servers spring to life, processing your request, pulling data, and generating a response. That’s inference—the real-time work that makes AI feel alive. But behind the scenes, it’s a massive, energy-hungry operation. And for every dollar OpenAI earns, a huge chunk goes straight to Microsoft’s cloud.
Analysts now believe OpenAI may be spending more on inference than it earns. “It’s like building a rocket to the moon, but the fuel costs more than the ticket price,” says Dr. Elena Torres, a tech economist at Stanford. “If this continues, the entire AI industry could be sitting on a bubble.”
The Microsoft-OpenAI Tango
So why does OpenAI pay so much to Microsoft? The answer lies in a complex web of investment, infrastructure, and exclusivity. Microsoft poured over $13 billion into OpenAI, securing not just a stake, but a privileged partnership. OpenAI runs almost all its models on Azure, Microsoft’s cloud, and in return, Microsoft gets a cut of every dollar OpenAI makes.
But it’s not just about money. Microsoft also sends a portion of its own AI revenue—like from Bing and Azure OpenAI services—back to OpenAI. Still, the leaked documents suggest the balance is heavily tilted. “Microsoft is the landlord, the power company, and the bank,” says tech analyst Marcus Reed. “OpenAI is the tenant, the tenant who’s burning through cash at an unprecedented rate.”
A Day in the Life of an AI Worker
Let’s zoom in on Maria, a data annotator in Nairobi. She’s one of thousands of workers around the world who help train AI models by labeling and reviewing data. OpenAI pays her employer, Sama, $12.50 per hour. After taxes and overhead, Maria takes home about $1.50. She works 10-hour shifts, reviewing snippets of conversations, flagging sensitive content, and making sure AI doesn’t go off the rails.
Maria doesn’t know about the spreadsheets or the billions. She just knows her job is hard, the pay is low, and the future is uncertain. But her work is the invisible foundation of the AI revolution—a revolution that’s costing more than anyone imagined.
The Ripple Effect
Governments are starting to notice. In Washington, D.C., lawmakers are drafting new regulations to monitor AI’s environmental and economic impact. “We can’t let a handful of companies control the future of technology while the rest of us pay the price,” says Senator Lisa Chen. Meanwhile, competitors like Google and Amazon are scrambling to build their own AI infrastructures, hoping to avoid OpenAI’s fate.
What’s Next?
The big question now is: Can OpenAI turn a profit? Or is the AI gold rush built on a foundation of sand? Some experts believe OpenAI will eventually find a way to cut costs, maybe by building its own data centers or developing more efficient models. Others warn that the entire industry could face a reckoning if the costs keep rising.
One thing is certain: the spreadsheet that started it all has changed the conversation. The AI revolution isn’t just about smarter machines. It’s about who pays for them, who profits from them, and who gets left behind.
Provocative Question:
If AI’s real cost is hidden in the cloud, who’s really in control of the future?
FAQ
Q: What is inference in AI?
A: Inference is the process of running an AI model to generate answers or predictions in real time. It’s what happens when you ask ChatGPT a question.
Q: Why does OpenAI pay Microsoft so much?
A: OpenAI uses Microsoft’s Azure cloud to run its models, and in return, Microsoft gets 20% of OpenAI’s revenue.
Q: Is OpenAI losing money?
A: Leaked documents suggest OpenAI may be spending more on inference than it earns, but the company hasn’t confirmed this.
Q: What are the implications for the AI industry?
A: High inference costs could lead to a bubble, where companies spend more than they earn, and only the biggest players survive.
Q: How does this affect everyday users?
A: If costs keep rising, AI services could become more expensive or less accessible to the average person.