Internal documents from OpenAI predict the company will lose $14 billion in 2026, according to a new report. Let that sink in: fourteen billion dollars. Not revenue. Loss.

This is the same company that recently closed funding rounds valuing it at over $150 billion. The same company whose CEO, Sam Altman, has been touring the world promising artificial general intelligence is just around the corner. The same company that's become synonymous with the AI revolution.

And they're reportedly burning cash faster than almost any company in tech history.

The business model problem is simple: running large language models is obscenely expensive. Each ChatGPT query costs OpenAI money—estimates range from a few cents to over a dollar per response depending on the model. Meanwhile, they charge $20/month for ChatGPT Plus. The math doesn't work unless most users barely use the service.

But here's where it gets interesting: OpenAI is reportedly now seeking another $50 billion in funding from Middle Eastern investors. That's not the behavior of a company with a path to profitability. That's the behavior of a company that needs to keep the music playing.

I've covered enough tech hype cycles to recognize the pattern. WeWork raised billions while losing money on every desk it rented. Uber subsidized rides for years hoping to achieve monopoly pricing power that never materialized. The question isn't whether OpenAI's technology is impressive—it obviously is. The question is whether anyone can make money at this.

The infrastructure costs are staggering. OpenAI's compute bills reportedly run into the billions annually. They're racing to build larger models that require even more computing power, creating a treadmill where costs scale faster than capabilities. And now competitors like DeepSeek are demonstrating you can achieve similar results with dramatically less computing power, potentially undermining OpenAI's entire moat.

Here's what the losses actually mean: is subsidizing every interaction you have with . They're paying enormous sums for cloud computing. They're poaching AI researchers with compensation packages that make engineers jealous. And they're doing all of this while generating revenue that doesn't come close to covering costs.