Is AI a bubble?

The short answer is yes. But not in the way most people mean it.

There are two kinds of bubbles in technology and finance. Greed bubbles and productivity bubbles. Understanding the difference matters if you want to make good decisions about AI over the next decade.

Two Types of Bubbles

A greed bubble forms when people are chasing money without creating real value. Capital flows faster than fundamentals. People make bets with cash and hope they exit before the music stops.

Crypto, specifically web3 rather than Bitcoin itself, is a clear example. The 2008 mortgage crisis with credit default swaps is another. In both cases, the system collapsed and very little durable infrastructure remained afterward.

A productivity bubble is different. It happens when capital floods into a space too quickly, valuations get ahead of reality, and many companies fail. But underneath the wreckage, real infrastructure and real capability are built.

The dot com era was a productivity bubble.

During that time, foundational work happened that shaped the next twenty years. Homes were wired for the internet. Web applications became viable. Software techniques were developed that made modern SaaS possible. When the bubble burst, many companies died, but the groundwork for massive growth remained.

Why AI Looks Like the Dot Com Era

AI today looks much closer to the dot com era than to a greed bubble.

Many AI SaaS companies will fail. Every software product is suddenly labeled AI, and a large percentage of those products simply do not work well enough to deliver value. Those companies will struggle or disappear.

That is normal.

What matters is that real infrastructure is being built underneath the hype. Models are improving. Tooling is maturing. Enterprises are learning where AI actually fits into operations rather than demos.

The companies that genuinely create value will survive and become very large. The rest will not.

OpenAI and the Strategy of Interdependence

OpenAI is aggressively tying other companies to its future success and failure.

Sam Altman is fundamentally a deal maker. I am not overly concerned about the circular money moving through the AI ecosystem right now, even though parts of it look strange. The strategy appears simple.

Make OpenAI too intertwined to fail.

Large language models are becoming a commodity. If ChatGPT disappeared tomorrow, it would be inconvenient for a few weeks, but organizations would migrate to Claude, LLaMA, or Gemini and continue operating. OpenAI knows this.

By embedding itself deeply into enterprise workflows, partnerships, and infrastructure, OpenAI increases the cost of failure. The circular capital looks less like speculation and more like business continuity insurance with investment layered on top.

Power Is the Real Constraint

The real limiter in AI is not chips. It is power.

There are more chips being produced than there is power available to run them efficiently. Data centers take quarters or years to bring online. Power infrastructure does not scale in months.

My sense is that the bubble bursts when demand for power significantly outpaces the ability to bring new capacity online. When that happens, return on investment timelines stretch, quarterly results suffer, and financial pressure cascades through the market.

Right now, demand for power only slightly exceeds demand for chips. That is why the system is holding.

Long Term AI Productivity Is Inevitable

Long term, AI will be more productive than the internet.

In the United States alone, ten thousand baby boomers retire every day. Generation Z is the smallest generation in modern history. Birth rates across the western world continue to decline.

Productivity is being removed from the economy whether we like it or not.

AI and robotics are the only credible path to replacing that lost output. Without them, the alternative is a prolonged economic slowdown in the 2030s or 2040s.

I am not particularly worried about job loss. Historically, the people who struggle are not those displaced by technology, but those who refuse to adapt to new ways of working. That pattern has repeated throughout history.