The AI Bubble

The AI Bubble

Over $3 trillion in data center spending is projected between 2025 and 2028. Every week, more than 800 million people use OpenAI's models. And yet, most companies report that AI has done nothing measurable for their productivity. That contradiction is the story of our moment.

Bull

Unlike the dot-com era, the companies at the center of today's AI boom are generating real revenue from real products. Adoption is accelerating: around 40% of American workers reported using AI at work as of late 2025, up 10% every year. The infrastructure being built today has lasting value regardless of which individual players are left standing.

Bear

OpenAI has consumed over $150 billion in investment to produce roughly $15 billion in 2025 revenue. Microsoft has already revised its AI sales targets downward. The MacroStrategy Partnership has called it "the biggest and most dangerous bubble the world has ever seen," estimating it’s 17 times larger than the dot-com at its peak.

Productivity Paradox

As of February 2026, 90% of firms reported no productivity impact from AI, even as executives continued projecting future gains. Another survey found that nine out of ten senior managers reported no measurable improvement from their AI initiatives. The gap between spending and results has not closed. If anything, it is widening.

Dot Com Comparison

The comparison cuts both ways. The parallels are obvious: inflated valuations, copycat companies, and infrastructure built ahead of demand. But today's major players are profitable in ways Pets.com never was. And history offers one enlightening note: the fiber optic cable buried during the dot-com boom outlasted every company that built it, powering the next generation of the internet. The builders lost. The builders of what came after won.

What Comes Next

Three paths: 1) a dramatic pop triggered by a credit event or high profile failure, 2) a slow deflation where valuations quietly correct and weaker players disappear, 3) a long plateau before true transformation arrives. Amara's Law (president of the Institute for the Future, a Silicon Valley think tank) suggests the third is most likely. We are almost certainly overestimating AI's short-run impact and will underestimate the long-run.

For investors, the priority is splitting companies with genuine AI advantages from those riding the hype. For executives, the productivity data is a warning: deploying AI without a clear, measurable use case is an expensive way to generate a press release. The people who laid the cables in 1999 went bankrupt. The people who built on top of those cables became billionaires. The AI boom may follow the same script, just with much larger numbers and much higher stakes.