Get the Gold, Not the Arrows

Everyone is building data centers... Lessons from the railway and the Internet booms... Big Tech is going into debt... Get the gold, not the arrows... AI is a disruptive force – and that's OK...

The machines are taking over...

At least that's how it felt when I (Dan Ferris) saw a chart posted on the social platform X...

Shared by the Kobeissi Letter, the chart shows that data centers under construction in the U.S. (including those under construction today) now exceed the value of America's incoming office buildings.

Office buildings are – or at least were – full of people. They have plenty of machines in them, but they also need a bunch of people to work all the computers and printers... coffee machines... everything. They exist to bring people together.

Data centers aren't built for people. The machines are stacked one on top of another... with just enough room for humans to squeeze between the servers when needed. These machines work day and night without direct on-site human intervention.

No smoke breaks. No complaints about pay. No vacations and no time off to go to the doctor.

Is there any doubt that this trend will continue? Not in my mind.

From instant communication to online shopping, from trading stocks to accessing medical records, the Internet has made people's lives much easier. Smarter AI software will provide even bigger benefits. And as regular Digest readers know, AI is fueled by those stacks of servers in data centers.

I expect more of that and a whole slew of products and services I can't presently fathom. I expect most folks will make a bunch of wrong predictions about AI like they did about the Internet. I remember one analyst insisting that people would never buy clothes online because they'd need to try them on first...

But I do believe AI will be another step-function change in our lives and daily routines, just as the Internet has been.

That doesn't mean building all these data centers is a good business move...

It's probably a terrible one.

In a recent (excellent) Substack piece, Simplify Asset Management's Mike Green recalled the Internet boom's great investments in fiber-optic data cables. From 2000 to 2022, Internet usage grew 1,000-fold... but telecom-services revenues (what you earned from owning all that cable) were cut in half.

That's what happens when a massive new capital investment causes supply to exceed demand. Some estimates near the bottom of the dot-com bust indicated that 90% of the fiber-optic cable installed to enable the Internet was still dark as of 2002.

As I've pointed out before, the biggest companies that laid all that cable – like Global Crossing and WorldCom, among several others – went bankrupt.

We've also seen this story long before the Internet boom. After the American Civil War, railroad companies laid 35,000 miles of track across the country.

The first transcontinental railroad was famously completed in 1869. But that was just one railroad. The area west of the Mississippi was vast, uncharted territory, ripe for settlement. Building the first railroad into a new area was thought to be a great business decision, giving the railroad owners an effective monopoly on long-haul transportation to that area (at least temporarily).

So they decided to build another one. The newly proposed Northern Pacific Railroad would run from Duluth, Minnesota to Seattle, Washington, connecting the Great Lakes region with the West Coast.

In 1870, a Philadelphia-based private bank called Jay Cooke & Company was chosen as the sole financier of the effort, but it struggled to sell its planned $100 million worth of railroad bonds. Cooke went bankrupt... And the Panic of 1873 was the result.

The panic decimated the companies that built the new transportation system, sending 89 of the country's 364 railroads into bankruptcy. Many investors were wiped out.

And yet, building all those railroads connected the country, just as the visionaries said it would.

Before the railroads were built, it could take six to 12 months to cross the U.S. on horseback or on foot. The railroads reduced that time to about seven days.

The railroads spread wealth across the U.S. by transporting building materials, consumer goods, and, of course, millions of people to settle in new, previously remote areas.

Expect the same story from data centers...

Enormous amounts of capital will flow into increasing data-center capacity as companies fear being left behind in the race for AI supremacy.

I've seen projections from $375 billion to $500 billion for the investment in new data centers next year.

If the data-center builders go the way of the railroads and the telecom services companies – and why shouldn't they? – then direct bets on those companies will wipe investors out.

Many smart investors will object to that prediction, since the biggest builders of data centers – the hyperscalers – are some of the world's greatest businesses: Microsoft (MSFT), Alphabet (GOOGL), Meta Platforms (META), and Amazon (AMZN).

It's unfathomable for them to go bankrupt... right?

Well, it's certainly highly unlikely now. Their core businesses generate plenty of cash, and they have tons of cash on their balance sheets... but they also have more debt than ever.

In fact, as a group, the four no longer have more cash than debt. After their latest flurry of data-center build-outs, they now have more debt than cash.

The debts are building... And the incentive to take on more – financing the race to build data centers – isn't going away.

In other words, no matter how good these companies look now, their financial conditions are deteriorating...

And the reason is all the money they've borrowed to build data centers.

Eventually, I believe the revenues from these data centers will drop, the same way the telecom-services revenues fell even as Internet usage skyrocketed.

So investors betting on the data-center builders will be disappointed. History tells us that... and with more certainty than it tells us anything else about the big, transformative technology build-outs.

And Amazon or Microsoft wouldn't have to go bankrupt to wipe out a whole generation of investors. All four of these tech giants are among the top 10 stocks in the S&P 500 Index, which is weighted by market cap and thus skewed toward its largest companies.

As I've warned before, nearly 72 million Americans have $12 trillion in their 401(k) accounts. They're overwhelmingly invested in index funds, most of which buy the S&P 500 or something very much like it.

If the market realizes Big Tech has spent billions on data-center overcapacity, millions of people are overly exposed to stocks that will take a big hit.

That's to say nothing of Nvidia (NVDA), the biggest S&P 500 stock of all, which makes much of its money by selling chips for these data centers. I expect Nvidia to perform like the top Internet equipment stock of the dot-com boom, Cisco Systems (CSO). It soared out of sight in early 2000, then crashed nearly 80% and still has yet to breach its 2000 highs.

And yet, just like with the railroads and the Internet, AI will transform our lives in untold ways...

I couldn't predict in advance how the Internet would change my life. But I certainly exploited the possibilities as soon as they arrived. This included meeting my wife online and moving to the West Coast to marry her, even though the company I worked for (Stansberry Research) is headquartered in my hometown of Baltimore.

It's the same with AI. Right now, the technology is in its earliest days. I don't know how it'll change my life and yours. And as investors, history has taught us that it's not necessary to know – and that relying on such a prediction to make you rich could be hazardous to your financial health.

As the old saying goes, "Pioneers get the arrows, settlers get the gold." Be patient. Be a settler. Get the gold, not the arrows.

Those fantastic businesses pouring money into AI should know that...

Microsoft didn't invent word-processing systems or spreadsheets. But it has the whole world addicted to Word and Excel.

Alphabet didn't invent Internet search, but people use the name of its flagship Google product as a verb for online searches.

Likewise, Meta didn't invent social media, and Amazon didn't invent e-commerce.

In their core businesses, they're settlers, not pioneers.

Unfortunately, with AI, they're all pioneers. They're building the new railroads across America and the world. You and I and our kids and grandkids will use these railroads to do... well, I have no idea what.

And however we turn out to use them, history doesn't suggest these AI pioneers will get a great return on their investment.

This doesn't mean AI is a bad thing...

Many folks are afraid that this technology will put lots of people out of work.

That's a typical reaction every time there's a big technological breakthrough.

In the early 1800s, textile workers in England – invoking a mythical figure named Ned Ludd – destroyed the new weaving machines they believed would replace them. And similar Luddites have been with us ever since.

Plenty of folks have similar fears about AI today. First Lady Melania Trump sounded a dramatic, cautious tone speaking at the White House yesterday:

The robots are here. Our future is no longer science fiction...

As leaders and parents, we must manage AI's growth responsibly. During this primitive stage, it is our duty to treat AI as we would our own children: empowering, but with watchful guidance.

"The robots are here" feels a lot like our observation about data centers overtaking office buildings in the U.S. We're building more homes for "robot" workers than human ones.

Now, it is undeniable that transformative new technologies can disrupt existing industries and displace workers.

But which do you have more of in your life: Friends and acquaintances who lost their jobs because of the Internet... or people who use it to conduct their business? For me, and probably for most of my readers, it's overwhelmingly the latter.

The Internet made it possible for millions of people to sell goods and services online. It has also fostered a publishing revolution, allowing writers all over the world to reach their audiences without a middleman.

That's how it tends to go... The technologies that disrupt existing industries create multiple new industries. Let's not forget the word "creative" in the phrase "creative destruction," an economic term for the very thing we're discussing.

The printing press displaced scribes and created whole subindustries for printing, publishing, editing, and other roles.

The automobile put the horse-drawn carriage out of business, and created an entire automotive ecosystem, from the production of raw materials like rubber and steel to a demand for people in sales, maintenance, and insurance.

It can be scary... But overall, we should cheer the arrival of new ideas with the potential to make life easier – and potentially much more meaningful – for human beings...

I have no reason to believe AI will be any different. We all know it's here to stay. And as we showed at the outset of this Digest, it's already displacing humans as the occupants of America's most feverishly constructed type of commercial real estate.

And that's just fine. Office real estate has been struggling for years, and the pandemic kicked that trend into a higher gear, with more folks working from home than ever.

Office workers aren't out of a job from that change. Like I did years ago, many of them have just changed where they work. And office buildings' old maintenance crews and security guards can wind up at data centers.

Perhaps you're familiar with certain folks who think AI and other technologies will turn large swaths of human beings into what one man has called "the useless class." This is the exact same argument as "new technology will put people out of work."

The day after you lose your job, you might not feel great. But humans are useful, not useless.

Don't waste your energy worrying about the machines taking over. Spend it believing in the never-ending ascent of man.

And don't bet your retirement savings on the data-center pioneers along the way.

New 52-week highs (as of 9/4/25): ABB (ABBNY), AutoZone (AZO), Barrick Mining (B), Ciena (CIEN), WisdomTree Japan SmallCap Dividend Fund (DFJ), DXP Enterprises (DXPE), FirstCash (FCFS), Alphabet (GOOGL), iRhythm Technologies (IRTC), JPMorgan Chase (JPM), Lumentum (LITE), Mueller Industries (MLI), Novartis (NVS), Omega Healthcare Investors (OHI), OR Royalties (OR), O'Reilly Automotive (ORLY), Ryder System (R), SSR Mining (SSRM), Travelers (TRV), Valero Energy (VLO), Vanguard S&P 500 Fund (VOO), and Wheaton Precious Metals (WPM).

In today's mailbag, a thumbs-up for a note in yesterday's mail about another factor for rising electricity prices (beyond AI data centers)... Do you have a comment or question? As always, e-mail us at feedback@stansberryresearch.com.

"Kudos to Subscriber Ken P. [in yesterday's mailbag]. He's totally on the mark. Florida residents are facing massive price hikes. The state insists on solar and battery as the future of our energy resource. The base load power needs to come from natural gas. It doesn't sleep, is very affordable and in massive supply. (And it's clean.) This is a national security issue." – Subscriber Thomas V.

Good investing,

Dan Ferris
Medford, Oregon
September 5, 2025