The Glass Thread

The Junk Bond Visionary

Gary Winnick had never laid a foot of cable in his life. What he had was something more valuable: the conviction that rules were for people who lacked imagination, and a Rolodex of the richest men in America. In March 1997, the former junk bond salesman announced he would build a fiber optic network spanning the globe. Within four years, his company would be worth $47 billion. Within five, it would be bankrupt.

Winnick learned his trade at Drexel Burnham Lambert, the legendary investment bank that revolutionized corporate finance by proving that high-risk, high-yield "junk bonds" could finance almost anything. His mentor was Michael Milken, the junk bond king who would eventually serve 22 months in federal prison for securities violations. But before the fall, Milken taught Winnick a lesson that would echo through the telecom boom: audacity scales.

The telecom opportunity was simple to explain and easy to sell. International phone calls cost dollars per minute. Data traffic was growing. Fiber optic cable could carry light pulses across oceans at nearly infinite capacity. Build the network, and the world would pay to use it. Winnick raised $750 million for Global Crossing in a matter of months, mostly from wealthy individuals who trusted him from his Drexel days.

"Winnick wasn't selling fiber. He was selling a story about the future — and for a brief, manic moment, everyone wanted to buy it."

The flagship project was AC-1: the Atlantic Crossing, a 14,000-kilometer fiber optic cable stretching from the United States to the United Kingdom and Germany. It would cost $850 million and take two years to complete. Traditional telecom companies had spent decades building such infrastructure. Winnick promised to do it faster, cheaper, and at a scale that would make the incumbents obsolete.

He had never managed a construction project. He had never run a telecom company. He had never operated a cable ship. But he had something those experienced executives lacked: the ability to raise staggering sums of money from people who believed that the old rules of business no longer applied. In the late 1990s, that was the only credential that mattered.

Bandwidth Barons

Winnick wasn't alone. A generation of executives bet everything on the same vision — and nearly all of them would crash together.

"Internet traffic doubles every 100 days."

This claim, repeated constantly by analysts and executives, was never true. Actual growth was closer to doubling every year — still remarkable, but nowhere near enough to fill the cables being laid. Watch the lie compound:

The Scale Defies Comprehension

Cathedrals Under the Sea

The North Atlantic, November 1998. The cable ship Tyco Decisive pitches through 40-foot swells, 600 feet of steel and purpose pressing westward. In its holds: 8,000 kilometers of fiber optic cable, coiled like a sleeping serpent. The crew works in 12-hour shifts, around the clock, feeding glass into black water.

The work never stops. Each day the ship lays roughly 200 kilometers of cable — feeding it over the stern at a precise speed, letting gravity and the cable's own weight carry it to the ocean floor. The fiber is thinner than a human hair, but the armored cable surrounding it is as thick as a garden hose. It has to survive the crushing pressure of the abyssal plain, the scraping of fishing trawlers, the chewing of sharks drawn to the electromagnetic field.

This is the AC-1 cable, Global Crossing's flagship project. Take the helm and lay it yourself:

When complete, AC-1 will connect New York to London and Frankfurt with capacity to carry 40 gigabits per second — more information per second than existed in all the world's libraries combined. The technology is called wavelength division multiplexing: multiple colors of light carrying different signals through the same glass strand. One fiber can carry 160 different wavelengths. One cable can carry 192 fibers. The math is staggering.

"They built cathedrals under the sea — engineering marvels that would outlast every company that paid for them."

But the ships keep coming. Global Crossing has three projects underway. Competitors have more. By 1999, there are dozens of cable ships crisscrossing the oceans, racing to lay fiber before the others. Each new cable makes the previous ones worth less. The executives don't care. They're building for a future where demand will be infinite. They're selling capacity before it exists, booking revenue before the cables are wet.

The irony: the engineering was perfect. These cables were built to last 25 years. The glass was flawless. The repeaters were robust. The routes were carefully surveyed. Every technical decision was sound. It was only the business case that was insane.

Irrational Exuberance

By the summer of 2000, Gary Winnick was worth $6 billion on paper. He bought the most expensive home ever sold in Los Angeles: a 60,000-square-foot Bel Air estate for $60 million. He filled it with Flemish tapestries and Old Master paintings. He threw parties where the guest list included presidents and movie stars. Global Crossing's market capitalization hit $47 billion. The company was three years old.

WorldCom was worth even more: $180 billion at its peak, making it one of the largest companies in America. Bernie Ebbers had assembled his empire through 65 acquisitions, gobbling up competitors at inflated prices, using his own stock as currency. The math only worked if the stock kept rising. It always had. He saw no reason it would stop.

The problem was simple: there was too much cable. Every company had built for a future where bandwidth demand grew exponentially forever. Instead, the rate of growth was slowing. Prices for capacity were collapsing — down 90% in some cases. The business models required scarcity. There was glut.

So the industry invented growth. Run the machine yourself:

The Reckoning

The dominoes fell fast. Once the first companies started failing, the entire sector's credibility collapsed. Investors who had believed in infinite growth suddenly saw the cables for what they were: expensive infrastructure with no customers. Tip the first one:

Dark Fiber

When the telecom industry collapsed, it left behind something unprecedented: millions of miles of unused fiber optic cable, buried underground and laid across ocean floors. In industry parlance, this was "dark fiber" — glass strands with no light passing through them, no signals, no customers. Just waiting.

What is Dark Fiber?

Fiber optic cables carry information as pulses of light. When a cable is "lit," data flows through it. When it's "dark," the glass is there — installed, tested, and functional — but no equipment is connected at either end. No light. No signal. No use.

By 2002, analysts estimated that 97.5% of all fiber capacity was sitting dark. The companies that built it were bankrupt or crippled. Their investors had lost everything. Their employees had scattered. But the glass remained. And glass, it turns out, doesn't rust.

"Glass doesn't rust, and cables don't care who owns them. The infrastructure outlasted every company that paid for it."

Vulture investors began circling. They bought fiber networks for pennies on the dollar — entire continental cable systems for the cost of a few urban buildings. Level 3 Communications, one of the few companies that survived the crash, began acquiring assets. So did private equity firms with long time horizons. They understood something the original builders had gotten wrong: the demand would come. It just needed more time.

The physical network was ready. It had always been ready. It was just waiting for the world to catch up.

Resurrection

The vindication came slowly, then all at once. The bankrupt dreamers had been right about the future — they just got the timing catastrophically wrong. Below: the fiber they left behind, one dot per bundle, as it slept in 2002. Drag the years forward and watch the world catch up:

The companies that had bought fiber on the cheap found themselves sitting on the most valuable infrastructure in the world. Level 3 Communications, which had survived by the skin of its teeth, grew to carry an estimated 70% of global internet traffic at its peak. In 2017, CenturyLink acquired Level 3 for $34 billion — far more than all the bankrupt telecoms had been worth combined.

The original investors and employees saw none of this. The founders who cashed out kept their mansions. The founders who didn't lost everything. But the physical infrastructure they built — the glass threads stretching across continents and ocean floors — became the nervous system of the 21st century economy.

They built the railroads. They went bankrupt. Someone else ran the trains.

Where Are They Now?

The Lesson

History has patterns, even if the participants never see them. In the 1860s, railroad barons laid track across America at a manic pace, fueled by government subsidies and investor euphoria. Most of them went bankrupt. But the railroads they built knit a continent together and enabled the industrial economy of the 20th century.

The Railroad Parallel

The telecom bubble offers both warning and comfort. Warning: the timing matters more than the vision, and the visionaries rarely survive their own revolutions. Comfort: even catastrophic failure can leave behind something useful. Sometimes the best way to build the future is to go broke trying.

Tonight, somewhere in the Atlantic, light pulses through glass strands laid by bankrupt dreamers a quarter-century ago. The cable ships have been sold. The executives have scattered or died. The money is long gone. But the glass endures, carrying Netflix queues and Zoom calls and cloud backups and everything else that constitutes modern life — a trillion-dollar gamble that paid off for everyone except the gamblers.

Frequently Asked Questions

Common questions:

Figures, dates, and fates as recorded in the public history of the telecom bubble; quotes as attributed in reporting of the era. The interactive exhibits are illustrative reconstructions — the ocean profile is stylized, the lit-fiber percentages after 2002 are directional, and no actual revenue was booked by the swap machine. The article itself reached you through glass laid by people who never got paid for it.