The AI Boom Isn't the 90’s Internet Bubble—It's the 90’s Global Fiber Glut, and That Changes Everything
There’s growing chatter in financial circles about a looming "AI bust."
The logic goes something like this: Trillions are pouring into data centers, GPUs, and AI startups. Big players like Nvidia, Microsoft, Google, and OpenAI are staking their futures on generative AI and machine learning. Valuations are sky-high. Everyone’s suddenly rebranding themselves as an "AI company." Sound familiar?
To many, it looks like the late 1990s all over again. The dot-com boom. The crash. Pets.com. The echo chamber is already prepping the obituary: "We’ve seen this movie before."
But I don’t think this is that movie.
I think we’re remembering the wrong chapter.
This isn’t the internet bubble. This is the global fiber glut — and if you understand what happened then, you’ll see why today’s AI buildout might stumble... but it won’t crash and burn.
The First Time the World Overbuilt
In the late 1990s, after the Telecommunications Act of 1996, a worldwide telecom frenzy began. Over $500 billion was invested globally to lay fiber-optic cables. We’re not just talking about cities. This was cross-continent and undersea. Miles of glass ran across oceans and deserts alike.
Companies thought they were building the railroad tracks for the next economic era—and that returns would be swift.
Then came the bust.
The dot-com collapse hit in 2000. Telecom companies like WorldCom, Global Crossing, and Qwest imploded. Demand for bandwidth didn’t materialize as quickly as expected. By 2002, reports showed that only 2.7% of installed fiber was being used. The rest sat dark.
That’s when the pundits came out. The fiber buildout, they said, had created "100 years worth of capacity."
It was meant as a punchline. A jab at the overbuilders.
But here’s what actually happened: We used it all up in about 15 years.
From Glut to Gold Mine
By 2015, thanks to innovations like dense wavelength division multiplexing (DWDM), every strand of glass could carry 10x, even 100x the data originally imagined. And boy, did we need it:
Streaming video
Smartphones and mobile apps
Cloud computing
Social media
Global e-commerce
The internet didn’t slow down. It exploded.
That massive overbuild became the foundation of the modern digital economy. Without it, there would be no Netflix, no Zoom, no AWS, no TikTok.
The AI Buildout Is Bigger — and Faster
What’s happening today with AI infrastructure is eerily similar.
Data centers are scaling globally. GPU clusters are going up in Singapore, London, Montreal, Frankfurt, and Texas. AI training consumes staggering amounts of compute and electricity. And we’re still in the early innings.
To skeptics, it looks like runaway spending on theoretical demand. But this time, something’s different.
AI is already unlocking productivity, automating workflows, analyzing mountains of data in seconds, and producing real economic value in sectors from healthcare to logistics to manufacturing.
Yes, some of this is frothy. Some valuations will come down. Some startups will disappear.
But the infrastructure won’t be wasted. It’ll be absorbed.
Just like the fiber.
The Real Risk: The Economy is Already Wobbling
Now let’s be clear: A short-term AI bust wouldn’t just hurt VCs and hedge funds.
The broader economy is already fragile:
U.S. federal debt is at 125% of GDP
Interest rates are high and unlikely to drop quickly
Consumer sentiment is weak
Trade disruptions and tariffs are weighing on margins
And unlike the late 1990s—when the economy was running on real growth and budget surpluses—today’s AI surge is one of the few forces propping things up.
If AI spending slows, the knock-on effects will hit cloud providers, chip makers, real estate developers, and countless downstream industries.
But the fundamentals of AI—like the fundamentals of fiber 20 years ago—are real.
So What Should We Expect?
Expect a shakeout. Expect layoffs and consolidation. Expect a few flameouts and a lot of "pivoting."
But also expect that:
AI infrastructure will accelerate the next wave of innovation
Businesses will find real cost savings and strategic advantages
The buildout will look oversized today but necessary tomorrow
Investors who can take the long view—and operators who build for resilience instead of hype—will be the ones who win.
Let’s Talk About It
Have you seen signs of AI irrational exuberance? Or are you seeing practical, valuable deployments?
Are you building your business to withstand a correction?
I’d love to hear what you’re seeing in your own industry. Let’s have a conversation.