If You're Worried About an "AI Bubble," You Need to Read This

In partnership with

I’m going to tell you a story.

I’m calling it “The Tale of the Bandwidth Bubble.”

During the late 1990s and early 2000s, companies invested half a billion bucks in high-speed networks — only to discover they’d spent their way into a bubble. And, like most bubbles, it ended badly — but (surprisingly) not as badly as most folks believe.

With companies projected to invest as much as $7 trillion in artificial intelligence (AI) projects between now and 2030, investors are worried we’re in bubble territory again.

I see parallels. But I see differences, too.

And successful Wealth Builders know that it’s the “what comes next” that matters.

In this ongoing series, I’m going to look at the “what’s next” for stocks … at storylines like the AI Era … at the health of the American economy … and at the wild cards to watch for.

I’m going to start with a serious look at AI and those “bubble” fears.

Investors want to handicap the present by comparing the big surge in AI stocks with the huge runup we saw in “Dot-Com” stocks 25 years ago.

But those “experts” are looking at the wrong things.

To really see what’s happening now — with the trillions being dropped on AI data centers, high-speed networks, hardware like chips and power sources — you have to study a specific slice of the Internet’s early days.

We have to look at the late-1990s “Bandwidth Bubble.”

Trust me when I say this, folks: I know what I’m talking about here — as a business reporter, I was a professional observer … with a ringside seat.

Feeding The Monster

It was 1999. The Dot-Com Era was four years old. I was a business reporter working in the Baltimore/Washington Region, which in those emergent days of the World Wide Web ranked as a Top 7 telecom center.

I drove a lot in those days. I had an hour-long commute — down to Baltimore City and back — every day. My assignments often took me down toward — and even into — Washington, D.C. And I did a fair bit of traveling for work and family to other parts of the country.

And everywhere I went, I saw one thing over and over and over again.

I saw telecom work trucks. With special trailers — each topped with gigantic “spools” wound with fiber-optic cable. Single spools. Double spools. Multi-man crews. Specialized tools. Everywhere you looked.

And everywhere you looked, telecom-and-utility crews were feeding that cable into underground ducts that ran along and beneath major highways — trenches, ducts and protective pipelines that also held conventional copper phone lines, power lines and other important networks of wire and cable.

And I-95 — the 1,900-mile highway that runs from Florida to Maine and is the key north/south thoroughfare for East Coast motorists — was like a real-life data study. Mile after mile, I’d see those trucks — and those crews — feeding that fiber-optic line into the ground.

As one expert said, I-95 became a “telecom trenching corridor.” Companies were trenching highways, railroads and rights-of-way.

It was like a ravenous giant monster lived underground and workers had to keep feeding it so it wouldn’t surface and wreaking havoc on businesses, villages and regular people. And fiber-optic cable was the only food that “monster” would eat. Those work crews were the appointed protectors of the rest of us — and they had to keep stuffing that cable into the ground — fearing it would never be enough.

Come to think of it, that metaphor — of the insatiable monster and the fear it would “never be enough” — is hit-the-bullseye accurate.

A Brand New Market Debuts — And Soars

While you can’t pick a precise point where the Internet went mainstream, it’s fair to say that 1995 was the year when the World Wide Web really took hold. But as a business reporter, I chronicled the boom – and warned of a bubble.

Amazon.com Inc. $AMZN ( ▼ 1.97% ) was launched in 1994 — and sold its first books online the following July.

And lots of companies, government agencies and everyday folks wanted websites of their own. From about 2,750 in 1994, the number of websites jumped to 23,000 the next year, 258,000 in 1996, an even million in 1997, 2.4 million in 1998, 5 million in 1999 and 17 million in 2000.

That’s aggregate growth of nearly 621,000% — or 6,200x.

In just seven years.

By itself, that’s stunning. But let’s layer in a bit of context. During that same stretch, the U.S. economy grew about 45% (or about 1.5x) and the Nasdaq Composite Index (a good proxy for the market-value growth) advanced 229% (or about 3.3x).

Check out the chart below.

During this run you had the rise of internet-service providers (ISPs) and the early broadband adoption. All the dot-com startups wanted faster connections. The Telecommunications Act of 1996 — the first update of U.S. communications law since 1934 — was super timely (or untimely, depending on your viewpoint) — supercharging mergers, igniting broadband investments and (eventually) leading to a concentrated media sector.

New telecom giants emerged — including WorldCom, Global Crossing, Level 3 and Qwest — and they and others laid hundreds of thousands of miles of fiber. Between 1996 and 2001, U.S. telecom companies collectively invested more than $500 billion in:

• That fiber-optic backbone.

• Switching and routing gear.

• Data centers and colocation facilities.

• And international cable systems.

Which brings us back to the “Bandwidth Bubble.”

Remember that $500 billion investment in telecom I talked about a few minutes ago?

Well, as much as 80% (about $400 billion) was bandwidth-related — routers, switching gear and fiber-optic cable.

Companies and investors believed this explosive growth would continue … heck, they were afraid of falling behind. They kept building. They kept investing. And they weren’t afraid to borrow and borrow and borrow to pay for it — a strategy that wiped out a lot of the companies that took that leap of faith.

The forecasts for bandwidth needs were massively overconfident — leading to massive overbuilding: By 2001, only a fraction of installed fiber was activated (referred to as “lit” in telecom parlance).

The Reckoning

That bubble burst in March 2000 — with the Nasdaq plunging 78% by October 2002. Scores of dot-com companies — most with no profits and many with no sales – were wiped away. The aggressive telecom players — including WorldCom — collapsed, too.

By the middle 2000s, an estimated 92% of all those fiber-optic networks were “dark” (unused).

Once again I saw this myself.

I actually helped cover Ciena Corp. $CIEN ( ▲ 1.01% ), a Maryland-based telecom-equipment firm that was positioning itself as a rival of heavyweight Cisco Systems Inc. $CSCO ( ▲ 3.14% ) —meaning it was a big beneficiary of the “Bandwidth Bubble.”

Another Baltimore Sun business reporter and I shared a major press association award for our coverage of the company.

The chart below tells you all you need to know about how hard many companies — and their stockholders — were hit when that bubble burst.

Most self-professed “students” of financial history end it there.

But there’s actually a happier ending.

It took years, but all that “dark fiber” eventually was “lit.” All that bandwidth was needed — in fact, more had to be built. Without that “head start,” the broadband and cloud-computing growth we saw in the 2010s may not have been possible.

And, as we know, the cloud was a table-setter for AI.

In short, the “Bandwidth Bubble” later fueled a “Bandwidth Arms Race.” The Baltimore/Washington Region became a hot zone for Internet tech, defense tech and (more recently) quantum computing because of that network capacity, its proximity to federal agencies and defense contractors, advanced research from Johns Hopkins and the University of Maryland, Aberdeen Proving Ground, Fort Meade and Naval Air Station Patuxent River (NAS Pax River), and the flight-testing center for the U.S. Navy — to list a few.

In short, it was a Good News/Bad News/Good News path, albeit one defined by a bubble that burst.

What does this tell us about the AI Boom as it relates to infrastructure?

That’s what I’ll look at next time.

Keep watching this ongoing series …

See you then,