Editor’s Note: Jim draws a striking parallel today between the AI data center boom and the 1873 railroad crash — and raises serious questions about whether companies like Equinix have gotten too far ahead of themselves. For a contrarian take on where the real AI winners are — outside the data center builders — see Jim’s AI Margin Rotation briefing.

I had my right knee replaced a few weeks ago, which means I am spending a lot of time at home. Although I don’t like being stuck at the house all day, I am glad that it has allowed me to catch up on my reading list.

Since I was taking a lot of pain medications in the immediate aftermath of the surgery, I thought it would be a good idea to start with the fantastical tale of “Dungeon Crawler Carl” by Matt Dinniman. But since I am an empiricist by nature, I had difficulty relating to the talking cat, exploding goblins, and drug-dealing llamas that populate this post-apocalyptic story.

However, I could discern the thinly veiled social message of the masses being subjugated to the whims of a select few. In that respect, that story was the perfect setup for the next book on my list, “1873” by Liaquat Ahamed.

The title of the book refers to the year that the global financial markets crashed as a result of excess speculation in railroad bonds. At that time, the technological superiority of railroads over all other forms of transportation induced otherwise rational investors to accept enormous risks that could not be clearly defined.

Runaway Freight Train

I feel we have entered a similar financial epoch due to the explosive growth of data centers being built to facilitate artificial intelligence (AI). We all know that AI will be a critical driver of the global economy in the years to come, but we cannot yet clearly define it in terms of scope and size.

Just as the railroad barons of the nineteenth century believed there was no limit to the economic benefits of building more infrastructure, so too are the tech titans of today in an arms race to create more data center capacity. And so far, they have had no problem raising money for projects that will cost billions of dollars and take years to build.

Also Read: “Chevron’s Power Play in Texas Isn’t a Done Deal Yet“

At some point, the data center market will become saturated. We probably aren’t there yet, but we may be getting close. When it does happen, we may see a repeat of what happened in 1873, this time due to overbuilding of a different sort.

Data Center REIT

As an investor, you can indirectly participate in data center growth by owning shares of a real estate investment trust (REIT) such as Equinix (NSDQ: EQIX). Equinix owns 281 data centers across six continents, serving over 10,000 customers.

Before the onset of 2026, EQIX delivered a total return of zero percent over the previous four years. This year, it is up 36 percent, as shown in the boxed area of the chart below.

Most of that gain has occurred since the onset of the war in Iran five months ago. Just as the outbreak of the coronavirus pandemic six years ago triggered a rush into tech stocks as a haven during a period of economic uncertainty, much higher fuel prices caused by the closing of the Strait of Hormuz have elicited a similar response this time.

Right now, data centers may seem to be a safe bet, just as railroads were thought to be in the years leading up to 1873. However, that is when risk is at its greatest since economic rationality is temporarily overridden by fear and greed.

Losing Steam

I do not doubt that AI will become a dominant force in our lives. However, I am not as confident that demand for data centers will continue to grow at an exponential pace for years to come.

If I am correct, then a simple way to profit from deceleration in demand for data centers is to buy a put option on a REIT such as Equinix. A put option increases in value when the price of the underlying security goes down.

Last week, while EQIX was trading at $1,035, the put option that expires in June 2027 at the $1,000 strike price could be bought for $100. For that trade to be profitable, EQIX must fall below $900 within the next ten months.

Bear in mind, EQIX was trading below $800 six months ago. But since the start of the war in Iran, it has taken off in response to Wall Street’s pivot towards AI infrastructure investment.

Switching Tracks

We won’t know how strongly Wall Street believes in the AI data center trade until the war in Iran is over for good. Until then, the put option trade described above is probably a losing proposition.

But once the war is over, Wall Street will reevaluate the multiples it is willing to pay for data center construction projects. At the same time, demand for data centers may decelerate as the efficiencies enabled by AI lessen the need for more processing capability in the future.

That does not mean that the AI data center market will disappear. To this day, many of the railroads built in the nineteenth century are still running on those same tracks.

However, there may be a day of reckoning approaching when the cost of expanding AI data center capacity exceeds its future financial benefit. And when that happens, data center REITs will be the first to feel it.

This is the same reasoning I applied to NVIDIA in January 2025: at some point, a sector’s market cap can’t be justified by its realistic future returns, and there’s no room for disappointment. I closed that position at +975% and rotated the proceeds into a different kind of company — not the AI builders, but the traditional businesses quietly using AI to widen their own margins. That rotation is what I’ve built the Personal Finance portfolio around this year. See the AI Margin Rotation briefing.