America’s data center REITs are crushing China (for now)

For over a decade now, I’ve been describing data centers as the physical infrastructure behind the digital economy. Because it’s true.

But this far into 2026, I have to acknowledge that data centers are no longer simply an economic or investment theme. They’re political, too, both on a national and global scale.

Across the U.S. – including in my hometown of Spartanburg, South Carolina – communities have been rising up against data centers for the water they consume… the electricity prices they affect… the noise they make… and the tax incentives they’re getting.

They’ve therefore become quite the hot topic among legislators as well. Politicians are recognizing more and more that voters simply don’t want these structures in their neighborhoods.

And many are taking steps to appease their constituents as a result.

President Trump recently warned about this NIMBY (not in my backyard) trend, calling it “backwards and poor.” China “could not be happier” with America’s anti-data-center movement, he said.

Certainly, both countries are in a race to achieve artificial intelligence (AI) dominance. And you probably saw at least some mention of Anthropic CEO Dario Amodei’s warning on Saturday that a “Chinese lead in AI would pose grave danger for the United States and the world.”

But America can’t beat China without data centers to back up our efforts. So we’re in a significant conundrum right now.

For the time being at least, we do have a clear edge in this regard. Our facilities are well-built, well-managed, and highly respected in the markets – with all the extra funding that entails.

But Trump is right that Chinese companies are working hard to support their own AI efforts. That shows in GDS Holdings Limited (GDS), one of China’s largest data center developers and operators.

It might not be up to par with our technological and economic capabilities just yet. However, that could very well change from here.

I first became aware of GDS through Gary Wojtaszek, former CEO of the now privately owned data center real estate investment trust (REIT) CyrusOne.

Gary, who joined the GDS board in 2018, was just re-elected to it in June. And since I follow him on social media, I got clued into the Chinese company as well.

GDS isn’t a perfect investment (as I’ll show shortly). But I can see the appeal, including how it booked 263 megawatts of digital space during Q2-26. That brings its first-half bookings to 471 MW – up significantly from last year’s total of 317.

Management thinks it can achieve 1,000 MW for 2026 all told, plus another 1,000 of customer reservations.

At last check, GDS had 1,589 MW in service, 579 under construction, and a 3,597 MW pipeline. Customers have committed to 2,047 MW, but only 1,289 were billable, leaving a 757-MW backlog. With that said, management believes it can approach 1.2 gigawatts (GW) by the end of the year.

Now, bookings aren’t earnings, of course. In order to generate cash flow, GDS has to first secure power, construct facilities, and deliver capacity to the customer.

Moreover, all that work requires an enormous amount of capital. So it’s important to note that while the data center company increased its 2026 organic capital expenditure forecast from ¥9 billion to ¥10 billion ($1.34 billion to $1.49 billion) – more than twice its 2025 spending – that’s still not enough to keep up with demand.

Here are some other numbers you should know, both good and less attractive:

GDS has ¥46.1 billion of gross debt and ¥31.2 billion of net debt. Its reported net leverage is 5.5x adjusted EBITDA, or about 4.7x after taking into account management’s pro forma adjustments. And interest coverage was 3.83x.

Essentially, the company’s balance sheet is manageable and capital remains available. However, there isn’t much room for either delays or spending slowdowns.

Then there’s customer concentration risk, with GDS’s three largest clients generating 77% of its first-half bookings. So if something happens to one of them, that would be a big blow.

In short, there’s a lot to like… and a lot left to improve upon.

In comparison, the U.S.-based Equinix (EQIX) remains the highest-quality data center operator around. Its global ecosystem creates high switching costs, so it tends to keep the customers it gets despite premium pricing. And it produced a 53% adjusted EBITDA margin in Q2.

Digital Realty (DLR), Equinix’s closest American competitor, ended the quarter with a $1.9 billion backlog of annualized signed-but-not-commenced rent. Plus, its leverage is low at 4.7x.

Both offer global diversification, investment-grade credit, and quality dividend payouts.

I’ll also point out Iron Mountain (IRM) even though it’s not even close to being a pure-play data center REIT. Only 13%–14% of its revenue comes from such facilities. Yet it still leased 110 MW through July, with more to no doubt come, offering growth right alongside the faithful dividend it pays.

Compared to these American plays, GDS is less diversified, more leveraged, and pays no dividend. It’s also exposed to Chinese regulatory and geopolitical risks – which are not something to take lightly (just look at Alibaba founder Jack Ma) – as well as variable-interest-entity risks.

It would, admittedly, be impressive if the company can turn that backlog it has into cash flow. Plus, GDS has a minority interest in international platform DayOne that’s intriguing. And though not a REIT itself, it has a pretty nice setup where it can profit from them.

For instance, last year, it sold a stabilized project to China’s NF GDS Data Center Infrastructure REIT for about ¥2.1 billion. But it still retained 20% ownership and continues to manage its assets.

If it keeps up with such setups, the company could recycle capital, strengthen private-market value, and reduce its balance sheet.

Regardless, it’s obvious what investors prefer year-to-date.

IRM and EQIX are up more than 35%, and DLR has gained nearly 22% – before counting their dividends. Whereas GDS is down almost 12%.

This hardly proves that America has won the AI infrastructure race, mind you. But it does show that investors currently place a much higher value on American REITs’ recurring distributions, better capital access, broader diversification, stronger governance, and more visible cash flows.

GDS could very well become a compelling investment in China’s AI buildout going forward if its enormous backlog becomes profitable. But until then, it’s not a sleep-well-at-night stock, and that’s a distinction that matters to Wide Moat Research.

Equinix, on the other hand, offers the widest moat in the data center buildout battles. Digital Realty is the strongest pure-play REIT. And Iron Mountain features the most interesting growth and income combination.

Most importantly, they’re all operating from a place of strength, not uncertainty. And that’s why investors prefer them year-to-date.

IRM and EQIX are up more than 35%, and DLR has gained nearly 22% – before counting their dividends. Whereas GDS is down almost 12%.

This hardly proves America has won the AI infrastructure race, mind you. But it does show that investors currently place a much higher value on American REITs’ recurring distributions, better capital access, broader diversification, stronger governance, and more visible cash flows.

GDS could very well become a compelling investment in China’s AI buildout going forward if its enormous backlog becomes profitable. But until then, it’s not a sleep-well-at-night stock, and that’s a distinction that matters to Wide Moat Research.

Equinix offers the widest moat in the data center buildout battles. Digital Realty is the strongest pure-play REIT. And Iron Mountain features the most interesting growth-and-income combination.

I have no problem recommending any of them at the right price.

As I mentioned in the beginning, the investment debate surrounding AI is only one part of a much larger conversation. There’s much more going on here than questions about which assets will work out and which may not.

As artificial intelligence consumes more power, capital, land, water, and computing capacity, society is left wrestling with whether it will ultimately enhance human productivity – or gradually diminish the value of human labor, judgment, and creativity.

It’s a question I plan to weigh in more fully in future articles. So stay tuned…

There’s lots more to discuss from here.

Happy SWAN investing!

Brad Thomas
Editor, The Wide Moat Daily