There is a number the AI industry really does not want you to think about too hard: 95%.

That is the percentage of organizations getting zero return on their generative AI investments, according to a report from MIT's Media Lab. Not a low return. Not a disappointing return. Zero. Nothing. The math equivalent of lighting money on fire and calling it innovation.

Now zoom out. US mega caps are expected to spend $1.1 trillion on AI infrastructure between 2026 and 2029. Total global AI spending will surpass $1.6 trillion. Microsoft alone burned through $35 billion in a single quarter on AI infrastructure. OpenAI was spending $1 million per day on Sora before they killed it.

Put those two facts together and the picture gets uncomfortable. The industry is pouring trillions into a technology that, for the vast majority of companies deploying it, is producing literally no measurable business value.

The Sora Lesson Nobody Learned

OpenAI's Sora shutdown last week is the canary in the coal mine. The video generation tool peaked at about a million users, collapsed to under 500,000, and was hemorrhaging money every single day it stayed online. OpenAI killed it not because the technology failed, but because it was going to look terrible on an IPO prospectus.

Here is the uncomfortable truth the industry keeps dodging: making AI models more capable costs exponentially more money for incrementally better results. The improvement curve is flattening. The cost curve is not.

Who Is Actually Making Money?

Nvidia. That is basically the answer. Nvidia made up about 7.3% of the S&P 500 at its peak, surpassed $5 trillion in market value, and is worth more than the GDP of every country except the US and China. AI-related stocks accounted for roughly 80% of all American stock market gains in 2025.

Everyone else? OpenAI is burning $14 billion a year. Anthropic is approaching $19 billion in annualized revenue but spending nearly as fast. Microsoft saw its share price drop 4% after earnings despite 18% revenue growth because investors looked at the AI spending and got nervous.

The Dot-Com Comparison Is Getting Louder

I am not saying AI is a bubble. The technology is real, the use cases are real, and some companies are genuinely transforming their operations. Two brothers just built a $1.8 billion company with literally two employees using AI. The potential is not in question.

But the current economics look a lot like 1999. Massive infrastructure spending. Astronomical valuations. A circular flow of investment where AI companies buy compute from cloud providers who invest in AI companies who buy more compute. And now a credible academic institution saying that 95% of the money going in is producing nothing coming out.

The internet survived the dot-com crash and became the backbone of the global economy. AI will probably do the same. But a lot of investors, a lot of startups, and a lot of corporate AI budgets are going to get vaporized between now and then. The math does not lie, even when the hype machine is running at full speed.