The $2 Trillion Case for an AI Bubble

Everyone wants to know whether artificial intelligence is a bubble.

My answer might surprise you.

Because the last major technology boom shares some striking similarities with what’s happening today.

A quarter century ago, investors poured hundreds of billions of dollars into what they believed would become the backbone of the digital economy.

And they were right. That technology changed the world.

Yet many of the companies at the center of the boom still got wiped out.

And researchers understood why long before most investors did.

The Fiber Boom’s Forgotten Lesson

Back in the late 1990s, investors couldn’t get enough of anything connected to the internet.

Telecom companies were raising enormous amounts of money. Fiber optic cable was being laid across continents and oceans. And Wall Street was convinced that the world would soon need vastly more bandwidth than existing networks could provide.

And to be fair, that prediction turned out to be correct.

Internet traffic exploded.

And the world eventually needed far more bandwidth than most anyone imagined. Especially once streaming and cloud computing arrived.

But investors made a critical mistake.

They assumed that the only way to meet this demand was by laying more fiber.

Meanwhile, engineers were working on a different solution.

While investors were focused on putting more cable in the ground, researchers at Bell Labs were figuring out how to send dramatically more information through the fiber that already existed.

In 2001, a  paper was published in Nature examining the theoretical limits of fiber-optic communications.

At the time, optical systems could transmit less than 2 terabits per second. But the authors of this paper found the theoretical capacity of a single fiber could reach roughly 100 terabits per second.

In other words, the same strand of fiber could eventually carry 50X more information than investors had assumed only a few years earlier.

That changed the math.

The world still needed more internet capacity, but it didn’t need nearly as much new cable as Wall Street expected.

And once that became clear, one of the biggest infrastructure booms in modern history crashed.

By 2002, roughly $2 trillion in telecom market value had disappeared. Twenty-three telecom companies had gone bankrupt. And the industry was carrying roughly $1 trillion in debt.

Yet internet traffic continued growing.

The problem wasn’t that investors misunderstood the future.

It’s that they underestimated how quickly the technology itself would improve.

That’s why I think about the fiber optic cable buildout whenever I’m asked if we’re in a bubble today.

Because just like back then, today’s AI boom has become an infrastructure story.

Microsoft, Amazon, Alphabet and Meta are spending hundreds of billions of dollars on chips, servers, networking equipment, power contracts and data centers. In fact, the largest technology companies in the world could spend nearly $700 billion on capital expenditures this year.

The assumption behind this incredible amount of investment is that AI demand will continue to explode.

And I suspect that’s true.

But there’s another side of the equation that investors aren’t talking about nearly as much.

Efficiency.

You see, the amount of computing power required for a task today might not be the amount required just a few years from now.

That’s why DeepSeek rattled the market so badly in early 2025. The company appeared to show that advanced AI models could be trained and run at much lower cost than many investors expected.

That announcement helped wipe nearly $600 billion from Nvidia’s market value in a single day.

And it should have served as a reminder to investors that efficiency can be disruptive too.

After all, that was the lesson of the fiber boom.

In the late 1990s, investors looked at exploding internet traffic and concluded the world would need endless amounts of new fiber.

They were right about the internet’s growth. But they were wrong about how much infrastructure would be required to support it.

We could be facing a similar situation today.

I still believe AI will be one of the most important technologies of our lifetime.

But if AI becomes dramatically more efficient, the industry may not need nearly as much infrastructure as investors currently expect.

And if that happens, the damage won’t be spread evenly.

The strongest companies will probably be fine. Microsoft, Amazon, Alphabet and Meta are not the debt-fueled telecom startups of the late 1990s. They have enormous cash flows, dominant businesses and real customers.

But not every company tied to the AI buildout has that kind of balance sheet.

Some companies are committing billions of dollars to projects that only make sense if AI demand unfolds exactly as expected.

Some power projects are being justified by AI demand forecasts that may or may not arrive on schedule.

And some equipment suppliers are being valued as if today’s spending growth will continue for years.

That’s where investors need to be careful. Because we’ve seen a version of this story before.

The internet changed the world.

But many of the companies that built its infrastructure never recovered.

Here’s My Take

The fiber boom helped build the modern internet.

It also wiped out many of the companies that built it.

And that may be the most important lesson for investors today.

Regards,

Ian KingChief Strategist, Banyan Hill Publishing

Editor’s Note: We’d love to hear from you!

If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to [email protected].

Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!