OpenAI's internal documents project the company will lose $14 billion in 2026, according to leaked financial forecasts reviewed by multiple outlets. That's not a typo: fourteen billion dollars, the largest annual loss in tech history by a company not named Tesla or Uber.

But here's the pitch: By 2029, OpenAI expects to be generating $100 billion in annual revenue with Nvidia-scale profit margins. It's either the most audacious growth story in modern business or the most spectacular case of wishful thinking since WeWork convinced investors it was a tech company.

Let's run the numbers. OpenAI generated roughly $3.7 billion in revenue in 2025, primarily from ChatGPT subscriptions and API access. To hit $100 billion by 2029, the company needs to grow revenue 27x in four years. That's not growth—that's a miracle.

For context, Microsoft took 20 years to grow from $3.7 billion to $100 billion in revenue. Google took 14 years. OpenAI is betting it can do it in four, in a market that didn't exist three years ago, competing against Google, Microsoft, Anthropic, and every other tech giant with a GPU budget.

The $14 billion loss isn't mysterious—it's compute costs. Training and running large language models requires staggering amounts of processing power, which means renting (or buying) thousands of Nvidia H100 chips at roughly $30,000 each. OpenAI's compute bill alone likely exceeds $5 billion annually. Add employees, research, infrastructure, and everything else, and you get a cash furnace that makes even venture capitalists nervous.

So where's the money coming from? Microsoft, primarily, which has invested over $13 billion in OpenAI and provides compute credits worth billions more. Other investors have poured in another $10 billion across various funding rounds. But even that war chest starts to look small when you're losing $14 billion a year.

OpenAI's bull case rests on two assumptions: First, that AI capabilities will continue improving at current rates. Second, that businesses will pay premium prices for access to those capabilities. If both hold true, OpenAI becomes the cloud computing platform of the AI era—think AWS for intelligence instead of storage.

The bear case? AI improvements plateau. Open-source models catch up. Enterprise customers balk at pricing. Regulatory scrutiny increases. Compute costs stay high. Any one of those could derail the path to profitability. All five happening simultaneously—not implausible—would be catastrophic.