Mispriced Capital Built Infrastructure

Why the worst telecom bubble in history accidentally wired the United States for the twenty-first century

How 1990s Internet Madness Built the Broadband Future

It’s hard to overstate how chaotic and irrational the internet boom of the late 1990s was, and how critical that frenzy turned out to be in wiring for the broadband age. In the span of a few manic years, dial-up internet went from a screeching novelty to a mass addiction. Stock investors threw money at anything with a “.com” or a fiber-optic cable. Billions of dollars were squandered on hare-brained ventures, and many fortunes went up in smoke when the dot-com bubble burst. By 2002, the Nasdaq had imploded, wiping out several trillion dollars of paper wealth. It was a disastrous bust for investors and companies alike. Yet buried in the wreckage of pets.com and WorldCom was an unlikely gift to the future: the physical infrastructure of broadband internet. The dial-up era’s strange economics and the dot-com bubble’s telecom frenzy together forced America’s phone and cable giants to lay the foundation of high-speed internet years earlier than they otherwise might have. This is the story of how an absurd, wasteful boom built something lasting despite itself.

Dial-Up Fever and the Great Phone Line Hustle

In the mid 1990s, getting online meant tying up your phone line and enduring the noisy handshake of a dial-up modem. What started as a slow trickle of people logging into CompuServ, a small local ISP, or AOL became, by the late 90s, a national obsession. Millions of Americans were suddenly camped out on dial-up internet for hours, checking their AOL email, browsing rudimentary web pages, or chatting on early messengers, all over ordinary telephone lines. Local phone companies (the Bell monopolies in each region) saw a surge in traffic unlike anything they’d experienced. A voice call averaged a few minutes; an internet session could last all evening. In areas where local calls were unmetered (which included most of the United States), this posed a unique problem: users could dial into the internet and stay on as long as they liked, all for the cost of their basic phone service. From the consumer’s perspective, it was an unlimited buffet of connectivity for one flat monthly fee. From the phone company’s perspective, it was a nightmare of network congestion and zero incremental revenue.

At first, the Bell companies tried to make the most of it. If people were going online all night and keeping the phone busy, why not sell them a second phone line? And indeed, that’s what happened. Households in the late 1990s eagerly paid for extra phone lines just for internet access, so Mom could still get calls while Junior tied up the modem. Telephone lines per household boomed, fattening the phone companies’ revenues in the short run. But this was a Band-Aid on a bigger shift. Americans’ love affair with being online was only deepening, and the strain on the old voice-centric network kept rising. Suburban phone switches started to get overloaded in the evenings. In some cities, the surge of internet usage was so sharp that it threatened to overwhelm local telephone infrastructure that had never been designed for continuous use by every household at once.

Worse (for the phone companies), a clever form of regulatory arbitrage emerged out of the dial-up craze. Thanks to rules encouraging competition, upstart local phone carriers (CLECs, in telecom jargon) could interconnect with the incumbents and get paid fees to terminate calls. Every time a Bell company’s customer dialed an internet provider served by a competitive carrier, the Bell had to pay that rival carrier a small fee for handling the call. These fees, known as reciprocal compensation, were meant to even out traffic between networks. But with dial-up, the traffic was all one-way: huge volumes of calls from Bell customers into the modem banks of ISPs often connected via competitive carriers. The result was a perverse incentive: some small carriers realized they could make a fortune by hosting ISP dial-up numbers. Every minute one of those AOL or local ISP sessions stayed live was money flowing out of the incumbent’s pocket into the upstart’s. The upstarts would then rebate some of that money to the ISPs or use it to offer ultra-cheap (even free) internet service. It was a classic arbitrage, exploiting a regulatory quirk to siphon off revenue from the big guys and effectively subsidize internet usage. For the incumbent Bell companies, this was an untenable bleed. Suddenly their most voracious dial-up users weren’t just annoying in tying up the network, they were directly costing the Bells real cash payable to competitors.

This bizarre situation couldn’t last forever. The incumbents naturally fought it in courts and commissions, and regulators eventually moved to curb the loopholes. But importantly, the dial-up arbitrage frenzy sent a clear signal to the phone giants: time was running out on the dial-up status quo. They could either watch their networks get choked and their profits drained by a zany patchwork of second lines, rival carriers, and free ISPs, or they could leap to a new model. That new model was broadband. High-speed internet access, delivered over DSL lines or cable modems, would be “always on” (no dial-up calls tying up circuits) and, crucially, could be billed separately from voice service. For the phone companies, getting customers onto DSL meant ending the dial-up free-for-all that was undermining their voice business. If a subscriber was on DSL, they weren’t making a metered phone call to an ISP anymore; no rival carrier could siphon off a termination fee. In one stroke, broadband could protect the phone companies’ turf and give them a new, lucrative service to sell.

So, by the late 1990s, the Baby Bells (SBC, Bell Atlantic, Ameritech, and the rest) reluctantly began overinvesting in digital subscriber line (DSL) technology to deliver faster internet over copper phone wires. This was something they had long been able to do (DSL had been invented years earlier) but had little incentive to deploy. Why rock the boat when dial-up users were still paying for second lines and per-minute long distance? The dial-up arbitrage mess changed that calculus. The Bells started rolling out consumer DSL in earnest around 1998, far earlier and faster than their original glacial timetable. They were economically forced to, or be crushed by the legal arbitrage of regulated voice lines. In effect, the dial-up boom forced the old phone system to evolve.

Cable Jumps Into the Game

The telephone companies weren’t the only ones facing the juggernaut. A parallel drama was unfolding with the cable TV operators. In the mid 1990s, cable companies were flush from years of monopoly pay-TV profits but had never provided two-way communication services. The idea that the local cable company could sell you internet access was novel. But as millions began flocking online via dial-up, the cable industry smelled both opportunity and threat. Opportunity because cable infrastructure, with some upgrades, could deliver much faster internet service than crackly old phone lines. Threat because if the phone companies somehow found a way to make their lines deliver broadband (which indeed DSL did), they might encroach on cable’s turf by delivering video or steal away young consumers’ attention entirely.

Around 1996, the first cable modem broadband trials kicked off. A startup called @Home Network partnered with major cable providers to launch high-speed internet over cable coax. Early cable broadband was blazingly fast for its day (initially 3–5 megabits per second, dozens of times quicker than a 56k modem) and had the advantage of being always-connected. By the late 90s, cable companies like Comcast, Time Warner, and Cox were pouring capital into upgrading their networks: stringing upgraded cable and fiber deeper into neighborhoods, installing new equipment in head-ends, and rolling out cable modems to customers hungry for speed. It was a massive investment, tens of billions, that fundamentally transformed the cable industry from a one-way TV pipeline into the two-way broadband behemoths we know today.

Why did cable operators move so fast? Part of it was the general gold rush mentality of the era: everyone felt they had to stake a claim in the internet boom or get left behind. Part of it was a fear that if they didn’t offer internet access, the phone companies would lock up that market with DSL. Cable also had an ace up its sleeve: unlike the telcos, cable broadband wasn’t regulated as a telephone service. The cable guys could offer internet access without having to share their lines with competitors or pay out weird reciprocal fees. They could charge what the market would bear and bundle internet with TV and (eventually) phone service. By 1999, cable broadband was available in many metropolitan areas, and it often blew DSL out of the water in performance. The race was on, and cable had a head start in the first lap of America’s broadband competition.

The existence of cable modems lit a further fire under the phone companies. If the arbitrage situation was the stick, cable competition was the carrot (or maybe a second stick) pushing telcos into broadband. No longer could the Bells dawdle with half-hearted trials, they had to invest heavily to catch up. Verizon (then Bell Atlantic) and SBC Communications, among others, ramped up spending to extend DSL to more neighborhoods and later to increase speeds. The telcos were rich, regulated giants, but in the freewheeling internet arena of the late 1990s, they suddenly looked like dinosaurs next to cable upstarts and agile new rivals. That reality led to some truly dramatic corporate moves, like AT&T (the long-distance giant, not a local Bell at the time) deciding in 1999 that it needed to buy an entire cable company (TCI, and later MediaOne) to get into the broadband game. AT&T’s foray was a strategic bust, it paid too much and ended up selling off those cable assets a few years later, but it underscores how urgent the broadband push had become by 2000. Every communications player saw that the future was high-speed internet, and they were willing to gamble the farm to avoid missing out.

The Dot-Com Bubble Builds the Backbone

While dial-up economics were shoving broadband into homes, Wall Street’s irrational exuberance was busy funding the backbone of the new internet at a breathtaking scale. The late 90s weren’t just the age of Amazon and eBay and goofy dot-com startups; they were also the age of telecom infrastructure mania. Investors, entranced by the exponential growth of internet traffic (real or imagined), opened the money spigots for any company that promised to build the highways of the Information Age. The 1996 Telecom Act had deregulated U.S. communications markets, theoretically to encourage competition. In practice, it encouraged a stampede of new entrants and projects. Suddenly there were dozens of fledgling telecom carriers laying fiber-optic cable across the country, building data centers, launching satellites, and stringing new networks in metro areas, all fueled by easy capital eager to ride the internet wave without having to bet on a profitless dot-com.

If you think today’s tech boom has large numbers, consider the late 90s: telecom companies raised and spent hundreds of billions of dollars on infrastructure in just a few years. They borrowed money like there was no tomorrow, and in a sense there wasn’t, because for many, tomorrow brought bankruptcy. Companies like Qwest, Level 3, Global Crossing, 360networks, and WorldCom raced to lay fiber and install network gear on the premise that internet usage would double, triple, quadruple ad infinitum. The mantra of the day was that data traffic was doubling every three or four months, an absurd exaggeration that nevertheless took on a life of its own. Investors bought it hook, line, and sinker. Who cared if a new fiber highway from Dallas to Los Angeles would run half-empty? Build it anyway, because surely we’ll need it soon enough. And so they built it, overbuilt it, in fact. By 2001, the United States had an astonishing glut of fiber capacity in the ground. One estimate later found that even four years after the bubble burst, over 80% of the fiber-optic strands laid in the late 90s were still completely unused (the famous “dark fiber”). It was as if the gold rush had left behind a ghost town of fiber-optic cables, waiting for a population of data that wouldn’t arrive until much later.

This boom-and-bust in telecom was spectacularly destructive to those involved. WorldCom, led by Bernie Ebbers, became one of the biggest frauds in U.S. history: its collapse in 2002 wiped out tens of billions in market value (and landed Ebbers in prison). Global Crossing, which spent wildly to string undersea fiber across oceans, went bankrupt in 2002 in a flash after its CEO cashed out stock options for hundreds of millions (just in time). At least twenty other telecom upstarts went belly-up around that period, victims of overambition, mismanagement, or simple bad luck when the capital dried up. Even venerable equipment makers like Lucent and Nortel, once riding high on orders for all this new network gear, imploded when those orders evaporated post-bubble. By 2002, telecom had become a dirty word on Wall Street. The party was over, and the hangover was epic. Investors lost not just money but faith; it would be many years before anyone dared to finance such speculative network projects again. In some circles the fiber mania was derided as the worst malinvestment in a generation.

And yet, for all that waste, the fiber got laid. The data centers got built. The internet exchange points and routers and gigabit links, however over-engineered in quantity for the year 2000, were now largely in place for the year 2010 and beyond, when they would finally be used to their full potential. In effect, the dot-com/telecom bubble forced an enormous upfront investment in America’s digital infrastructure. It was irrational and unsustainable as an investment thesis, but as infrastructure policy (unintended), it achieved what no government program or cautious private market likely would have: a near-ubiquitous fiber foundation for future broadband. When the dust settled, many of these bankrupt networks were sold for pennies on the dollar to more stable players. The new owners, often the surviving Baby Bell companies or well-capitalized firms that waited out the carnage, found themselves in possession of state-of-the-art fiber networks at bargain prices. They could then use this excess capacity to upgrade services cheaply. Bandwidth in the 2000s became dirt cheap compared to the 90s, thanks to the glut. That, in turn, made it economically feasible to launch data-heavy services like video streaming, online gaming, and cloud computing, which might have been cost-prohibitive if bandwidth prices stayed sky-high. In short, the bubble subsidized the bandwidth bonanza that powered the next two decades of internet innovation.

From Bust to Boom (for Consumers)

By the mid 2000s, the picture had flipped. Broadband was no longer a niche experiment; it was the new normal. In the year 2000, only a few percent of American households had broadband internet (most were still on dial-up). But as the new networks lit up, adoption skyrocketed. Cable and DSL lines spread to virtually every suburb and city. By 2005, more than half of U.S. internet users had a high-speed connection at home, and dial-up was in steep decline. The ironies were rich. The early 2000s were a bloodbath for telecom investors and dot-com dreamers, yet consumers were getting more value than ever. If you were a regular American web user, the post-bubble world was a wonderland: suddenly you could get a fast, always-on internet line at a reasonable price, often from the very same phone company or cable company that, just a few years prior, had been dragging its feet. The competition between DSL and cable modem service meant providers couldn’t gouge too heavily, each had to keep up or lose customers. The end result was that U.S. broadband prices kept trending downward and speeds kept going up throughout the 2000s, even as the companies providing it consolidated and regained their footing.

Crucially, the infrastructure was now largely built. The heavy lifting (and spending) of laying fiber and upgrading networks had been done during the boom. After the bust, plenty of physical capacity was sitting around, waiting to be lit up and used. The marginal cost to turn on service for new customers or boost network speeds was low. So telcos and cable companies, once they recovered from their debt hangovers, could profitably serve a growing base of broadband subscribers without immediately hitting a capacity wall. They had excess runway, thanks to all that overbuilding. In fact, one could argue the U.S. likely wouldn’t have achieved the scale of broadband adoption it did by the late 2000s if not for the bubble’s overshoot. A rational, incremental build-out of fiber and high-speed systems might have been slower, more geographically limited, and more expensive per user. The bubble era effectively front-loaded a decade or more of infrastructure development into just a few wild years.

Even some of the more bizarre dial-up artifacts ended up leaving long-term marks. Remember those second phone lines and the flat-rate local calling that nurtured the dial-up culture? They helped solidify the idea that internet access should be unlimited, unmetered, all-you-can-eat. Americans came to expect that you pay a flat monthly fee for internet and then use it as much as you want. That expectation carried over to broadband. Unlike some countries where early broadband was metered or capped tightly, in the U.S. the norm became unlimited usage (within reason). This in turn encouraged all sorts of high-usage online services to flourish, from Napster and YouTube to Netflix streaming and massive multiplayer games. Culturally, the freewheeling dial-up days set the stage for an internet ethos of abundance over scarcity, a notion that would have been hard to foster if we’d all been counting minutes online.

Blunt Lessons and Lasting Legacy

It’s easy, and not wrong, to ridicule the late 1990s internet bubble as a period of collective insanity. There were inane business plans, accounting scandals, absurd valuations, and a lot of smart people got burned badly when reality reasserted itself. But it’s also important to acknowledge that the frenzy did achieve something tangible and profound for society. In the pursuit of quick riches, the dot-com and telecom speculators of the 90s ended up accidentally laying down the tracks for the broadband future. They forced entrenched monopolies to innovate faster. They pushed an entire industry into the next technological paradigm almost overnight. The process was hugely wasteful, yet the waste was mostly in the form of overpaid stocks and redundant companies. The concrete outputs, the fiber in the ground, the upgraded cable systems, the nationwide data centers, were very real and very useful once the madness passed.

No sane executive in 1995 would have said, “Let’s risk bankrupting the company to install way more network capacity than we can use.” But by 1999 many were effectively doing just that under investor pressure, or because they feared (or hoped) that if they didn’t, a competitor would. The competitive and speculative dynamics of the bubble created a kind of arms race of overinvestment. That’s a terrible thing for those footing the bill, but a boon to the broader economy that inherits the infrastructure. In the decades since, America’s internet ecosystem has leaned on the surplus capacity built in that era. Even today, when you stream a 4K movie or join a Zoom call, there’s a decent chance some of the data is traveling over fiber routes originally installed during the 1998–2000 binge, or through an exchange point set up during that time. The players and logos have changed, WorldCom’s lines became MCI’s lines, now folded into Verizon, for example, but the physical cables don’t care who went bankrupt to put them there.

There’s a blunt truth here: market manias, for all their waste and excess, can sometimes accomplish what sober planning won’t. The United States in the late 90s didn’t have a coherent national broadband strategy or successful federal program to build internet infrastructure. What it had was a frenzy of private capital and competition, which in a chaotic way achieved the rapid diffusion of broadband. Of course, this path is nobody’s idea of an optimal strategy. It created winners who perhaps didn’t deserve to win and many losers who definitely didn’t deserve their fate. It led to a very uneven rollout (some rural areas still got left behind, since even the bubble had its blind spots). And it fostered a duopoly of cable and telco that we still grapple with in terms of competition and service quality. But the second-order effect of the bubble, an American internet backbone far ahead of schedule, paid dividends for years, and is still paying them. The societal benefit of a connected world is enormous, enabling new industries and innovations that those 90s investors couldn’t even dream of: social media, smartphone apps, cloud services, etc.

In hindsight, the dial-up arbitrage schemes and the dot-com telecom bubble look like textbook cases of stupidity and hubris in business. And they were. Yet, history has a sense of humor. Those follies paved the way for a new era of communication. The dial-up era forced broadband to happen; the bubble-funded pipes ensured broadband could scale. This doesn’t redeem the bad decisions or excuse the fraudulent ones, but it does illustrate the complex, often counter-intuitive ways that progress unfolds. Sometimes, it takes a mania to jolt an entire system into its next phase. The 90s internet boom was just such a big jolt, and broadband was the enduring gift left under the tree after the wrapping of hype was torn off.

As a result, by the time the 2010s rolled around, Americans were living in a broadband world largely constructed during that insane rush. What was irrational on a micro level turned out to be transformative on a macro level. In the tale of dial-up and dot-com excess, there’s a caution and a silver lining: Capital can be horribly misallocated in the shorter run and still move society forward in the longer run. The key is recognizing the difference between investor outcomes and infrastructure outcomes. Investors from the 1990s boom have plenty of scars to remind them of what went wrong. The rest of us, however, are still enjoying the digital highways they paved, often without even knowing it. And for all the pain and absurdity of that era, that enduring legacy counts as a win for society: one built, improbably, on the ashes of irrational exuberance.