Global Crossing Had a Massive Backlog Too

2 days ago

3 min read

The Buildout Series · Part 1 of 3

Demand for bandwidth in 2000 was real. Common shareholders still lost everything. The problem wasn't the business. It was the timeline. I know I'm going on about this but... I traded through the whole DotCom period and this needs some deeper analysis.

The Read

If you started investing in the last fifteen years, you may have only ever seen a bull market. So here's a history lesson worth having in your back pocket this week.

Demand for bandwidth in 2000 was real. The internet was coming, everybody knew it, and companies like Global Crossing raced to wire the planet with fiber. Global Crossing was worth about $47 billion in 1999. It had customers. It had long-term capacity contracts sold years in advance, its own version of a backlog. Then on January 28, 2002, it filed for bankruptcy with $12.4 billion in debt. At the time it was the fourth-largest bankruptcy in U.S. history. Common shareholders got nothing.

The internet got built anyway. It changed everything, exactly like the bulls said. Global Crossing's shareholders just weren't around to collect. They invested in a good business that didn't run its business well.

What Actually Sank Them

It wasn't fake demand. It was building with borrowed money on a timeline the debt couldn't wait for.

Write that on a Post-It note and stick it on your monitor. Interest is due every month. The revenue from a giant infrastructure buildout shows up over years. When the gap between those two gets too wide, and the lenders stop lending, it doesn't matter how right you were about the future.

By late 2001, Global Crossing hadn't signed up enough customers to cover its costs, some of its customers were in trouble themselves, and the investment community slammed the door on new funding. Game over, with the network about 85% built.

Now Look At CoreWeave

CoreWeave has a revenue backlog of about $104 billion. That's real, signed business, and about 41% of it is scheduled to be paid through 2028. Sounds bulletproof. But a backlog is a promise of future revenue. It isn't cash in the bank today.

Here's the mismatch. CoreWeave carries about $35.6 billion of debt plus roughly $16.3 billion of lease obligations, around $52 billion total. Interest expense was $640 million in the second quarter and is guided to roughly $900 million in the third. Meanwhile, the company posted a GAAP operating loss of $49 million that quarter. The operating line didn't cover the interest bill.

And it has to keep building. Capex guidance this year is $35 to $39 billion, against revenue of about $12.4 to $13.2 billion. So CoreWeave needs to raise something like $35 billion a year to build the capacity its backlog requires. That's tough in any market. It's a lot tougher when your bonds are trading like this one.

Bar chart comparing CoreWeave's 4.2 gigawatts of contracted power with 1.5 gigawatts active at the end of June

The backlog only turns into cash as data center capacity actually comes online. CoreWeave had 1.5 gigawatts of active power at the end of June, against 4.2 gigawatts contracted. That gap is the whole story. Are we having trouble building data centers? Yup. Power, permits, local opposition. Again: a timeline the debt can't wait for.

The Bond Market Already Noticed

CoreWeave's 2032 notes went from yielding in the low teens at the start of October to above 16% this week, with the bonds trading around 78 cents on the dollar. Credit default swaps, basically insurance against a default, are near 1,000 basis points. That's a very fast repricing.

Here's the key point, and it's the one most people get wrong. The bond market isn't saying AI demand is fake. AI demand is massive. It's saying CoreWeave's access to capital just got a lot more expensive. Lenders aren't worried about the GPUs already running. They're worried about the buildings that aren't powered yet. You need the buildings to pay the bills. Math.

Hans's Take

Yeah, I know, CoreWeave has the backlog, Microsoft as a customer, and Nvidia in its corner. Global Crossing had a great story too. The stock market doesn't pay you for being right about the future. It pays you for owning the company that survives long enough to collect. In Part 2, I'll walk through what rhymes between the fiber boom and the AI buildout. In Part 3, what's actually different. Trade smart, stay hedged.

Educational content only. Not financial advice. Options involve risk and are not suitable for every investor.