Alphabet Just Failed the 1873 Test

One morning in September 1873, a telegram reached Andrew Carnegie at his summer cottage in the Allegheny Mountains. Jay Cooke's bank had failed.

"Almost every hour after brought news of some fresh disaster," he wrote later.

Carnegie was 37, and he had a new steel mill going up outside Pittsburgh. He'd named it after Edgar Thomson, the president of the Pennsylvania Railroad, and both Thomson and the railroad's vice president, Tom Scott, owned shares in it.

Scott was also the man who'd given Carnegie his big break. Back in 1853, he'd hired Carnegie, then a 17-year-old telegraph operator, as his clerk for $35 a month.

As the panic spread, Scott telegraphed Carnegie to meet him in Philadelphia, "without fail."

Scott had been building a railroad of his own, the Texas Pacific, and Carnegie had paid $250,000 in cash for a stake in it.

Now a big loan for the railroad had come due in London. The bankers would renew it only if Carnegie signed on, too.

Carnegie said no.

He wrote later that he was asked "whether I would bring them all to ruin by refusing to stand by my friends," and he called it "one of the most trying moments of my whole life."

He told Scott he'd done his best to talk him out of starting a railroad before he had the money to build it, because "thousands of miles of railway lines could not be constructed by means of temporary loans."

The next morning, Carnegie rode to New York with the group in a private railroad car.

Along the way, one of the railroad's vice presidents looked around the car and said there was only one sensible man in it. The rest were "fools."

The sensible one was Carnegie, who "had paid for his shares and did not owe a dollar."

His Name Wasn't on One Dollar of Their Paper

Not long after that ride, the crash caught up with Scott and Thomson's Texas Pacific venture.

In Pittsburgh, the president of one of Carnegie's banks told his board it was "simply impossible" that Carnegie wasn't caught up in it too. He suggested the bank stop lending Carnegie money against what his customers owed him.

Carnegie took the first train to Pittsburgh and told everyone his name "was not upon one dollar of their paper or of any other outstanding paper."

In plain English, he hadn't co-signed a single IOU outside his own business.

It worked. Before long, in his words, "our credit became unassailable."

From then on, whenever money got tight, lenders offered him more of it, not less.

Back when the panic first hit, he'd had to stop building the new mill for a while. Some of the people who'd invested in it couldn't keep up with their payments, so Carnegie took over their stakes and paid each of them back in full.

That's how he ended up in control of the company.

Carnegie Walked Away With $225 Million

In 1901, J.P. Morgan's group bought Carnegie's steel company for $480 million, and Carnegie's share came to about $225 million.

Not bad for a kid who came over from Scotland at 12 and started out as a bobbin boy in a cotton factory for $1.20 a week.

My parents were immigrants, too. They worked hard their whole lives and never built much wealth.

Alphabet Came Up $9.1 Billion Short

On Monday, in Big Tech Is Borrowing Like It's 1873, I said the first thing I'd check this week was whether each tech giant's spending ran ahead of the cash coming in, and how it covered the gap.

That's what I mean by the 1873 test.

Alphabet, Google's parent company, reported Wednesday night. Over the three months through September, its business brought in $42.3 billion in cash, and it spent $51.4 billion on data centers and equipment.

That left it $9.1 billion short, after coming up $5.9 billion short the quarter before.

So how did it cover the gap? It borrowed about $29 billion in those three months.

A year earlier, Alphabet spent $11.5 billion buying back its own stock over the same stretch.

This time, it bought back none.

Its long-term debt has gone from $46.5 billion at the end of last year to $125.1 billion. And it told investors to expect a lot more spending in 2027.

In 1873, Carnegie stopped building.

To be fair, Alphabet is nowhere close to broke. It ended September with about $265 billion in cash and investments, though a big chunk of that is SpaceX stock it can't sell yet.

But it's now paying for its data centers partly with borrowed money.

The shares fell more than 12% on Thursday, their worst day since Google went public in 2004.

Investors had other complaints, too, like slower growth in search ads and profits that came in below forecasts.

Microsoft passed. It took in $58 billion of cash from its business, plenty to cover the $39.4 billion it spent on data centers and equipment. Its stock held up far better than Alphabet's.

Then Amazon reported Thursday night, and it came up even shorter.

Its business brought in $42.4 billion over the same three months, and it poured about $76 billion into data centers, warehouses and equipment.

That left it about $33.5 billion short in a single quarter. It borrowed about $48 billion.

Its free cash flow over the last 12 months is negative $42.4 billion. A year ago, it was a positive $14.8 billion. Its stock fell sharply after hours.

NVIDIA Signed On for Up to $105 Billion

Today, some of the companies selling to the AI giants are doing exactly what Carnegie refused to do. They're co-signing for their customers.

NVIDIA owns a piece of OpenAI, the way Carnegie owned a piece of Scott's railroad. In February, it agreed to invest $30 billion.

But in August, it went a step further. It agreed to stand behind OpenAI's 20-year lease on a giant data center campus in Ohio, which means that if OpenAI ever stops paying, NVIDIA is on the hook for part of the rent and power bills.

NVIDIA has also signed a $6.3 billion order to buy whatever computing power CoreWeave can't sell, through April 2032. That's the same CoreWeave I told you about on Monday, the one that expected about $4 of interest for every $1 of operating profit.

Carnegie wouldn't put his name on one dollar of his own mentor's paper.

NVIDIA put its name behind a customer's lease, for up to $105 billion.

Harry Seldon, who writes our technology issues, called NVIDIA's latest quarter "extraordinary financial strength" in Thursday's issue, and he's right. But if OpenAI or CoreWeave stumbles, NVIDIA can lose twice, once on the chips and again on the promises.

Now look at GE Vernova, which makes the gas turbines that power plants run on and the electrical gear AI data centers need.

Its customers are paying just to hold a place in line.

On Wednesday, its finance chief said its cash flow got a $5.3 billion lift last quarter. Most of it came from down payments on new orders and those reservations.

Its prices are climbing, too, and it expects gas turbines ordered this year to bring in about 50% more than last year's, for the same amount of power.

And it just doubled its stock buyback plan to $20 billion.

Find Out Who's Carrying Whom

Here's your one move this weekend.

Pick an AI stock you own, or the one you've been itching to buy, and pull up the transcript of its latest earnings call. That's the call it holds with Wall Street analysts after every quarterly report. Search the company's name plus "earnings call transcript," and you'll find it free.

Then search the page for two sets of words.

1. Down payment, deposit, reservation, prepayment. That's customers paying before they get what they bought.

2. Invest, finance, backstop, guarantee. Next to a customer's name, they're a sign the company is putting its money or its name behind the people buying from it.

If you find plenty of the first and very little of the second, Carnegie would've liked that company.

If it's the other way around, use Carnegie's rule for co-signing anything.

He said to ask two questions first. Could you pay the full amount "without inconvenience"? And "Am I willing to lose this sum for the friend for whom I endorse?"

Only put in as much as lets you answer yes to both.

If it were me, I'd lean toward the companies holding their customers' money, because when lenders pull back, those companies have already collected part of the bill.

What Premium Members Get Today

Everything above is free and yours to keep.

If you ran the check and want to know which companies we actually own, that's the part I save for Premium Members.

Since launching Premium, we've closed 14 trades. All 14 made money, with an average gain of 84.5%, against 40.18% for the S&P 500 over the same periods.

In case you missed it earlier this week, today's Premium section is a review of one of our critical-minerals positions that's gone against us, checked against the exact conditions we published the day we bought it, along with what would make us sell.

When a pick isn't working, you deserve a straight answer, and Premium Members get one below.

Double D

🔓 Premium Content Begins Here 🔒