Latest Earnings Show ‘A Tsunami Of AI Demand’ Hitting Data Centers

Artificial intelligence continues to boost data center demand in what could end up being a record leasing year for the sector.

The second quarter saw hyperscale giants like Microsoft and Google continue their massive investments in data centers to support AI, driving a streak of massive lease deals that may just be a preview of what’s to come.

But even as the world’s largest tech companies plan to further increase capital expenditures in the months ahead, data center REITs Digital Realty and Equinix may not be in a position to fully capitalize on the opportunity. 

AI Fuels Rising Capex, Record Leasing For Hyperscalers

Led by Microsoft and Google, major cloud providers and tech giants are ramping up data center capex as they race to build out the infrastructure to support AI applications that are increasingly central to their business models.

Microsoft’s overall capex rose to $10.7B last quarter, a nearly $3B jump from the preceding three months that the company attributes largely to spending on data centers and other digital infrastructure like servers and networking. On its quarterly earnings call last week, the Seattle tech giant’s leadership indicated that capex would continue to rise every quarter through the end of 2024 as the company attempts to stay ahead of a growing wave of AI-driven demand. 

Google parent company Alphabet also intends to crank up data center spending in the months ahead to support its wide-ranging AI initiatives. Although the company’s overall capex actually dropped slightly to $6.89B in the second quarter, Google’s leadership attributed the lower-than-expected figure to what it termed “delays in certain data center construction projects,” with significant increases in capital expenditures expected over the coming 18 months. 

“We expect elevated levels of investment in our technical infrastructure increasing through the back half of 2023 and continuing to grow in 2024,” Alphabet Chief Financial Officer Ruth Porat said on its earnings call last week. “The primary driver is to support the opportunities we see in AI across Alphabet.”

The quarterly numbers are just the latest evidence that Big Tech’s AI arms race has begun to dramatically transform the data center landscape over the past three months. A report released last week by investment bank TD Cowen outlined what it termed “a tsunami of AI demand” hitting the data center leasing market, with hyperscalers desperate for new capacity potentially driving leasing numbers to new highs this year. 

According to the report, around 2.1 gigawatts of data center leases were signed in the U.S. in the preceding 90 days, a remarkable figure considering that the entire U.S. third-party data center market has less than 10 gigawatts of capacity. The report said some data center providers have seen their leasing pipelines more than double in 90 days, a dramatic demonstration of the rapid pace at which hyperscalers’ AI needs are driving the market.

Microsoft and Google have led the charge on this surge in leasing demand. The two tech behemoths accounted for 1.8 gigawatts of leased capacity through a handful of enormous deals with third-party providers. 

Microsoft alone took down more than a gigawatt of capacity, signing a 420-megawatt deal in Leesburg, Virginia, a 360-megawatt deal in Dallas and a 300-megawatt deal in Chicago, according to TD Cowen. Meanwhile, Google signed a massive 600-megawatt agreement in Texas. 

“Not only have the number of [signed or in-process] hyperscale deals increased, but also the size of these deals has increased meaningfully,” the report’s authors said. “Amid this torrent of demand, we expect 2023 to set a new record for data center leasing.”

More deals are expected in the months ahead, according to TD Cowen, with multiple hyperscalers actively looking for data center providers to develop facilities for deployments north of 500 megawatts. 

Despite AI Demand, Data Center REITs Have Yet To Make A Splash

The campuses to host these massive hyperscale deployments are being developed by private data center providers, while data center REITs Equinix and Digital Realty have remained on the sidelines — at least for now. 

Digital Realty has not landed any of the major AI-driven hyperscale deployments over the past quarter, and it did not sign any deals larger than 100 megawatts at all, even as smaller AI applications boosted its overall pipeline and contributed to strong leasing numbers, it said on its earnings call. But CEO Andy Power told analysts that the company is interested in landing these large-scale development deals and has been pursuing them.

At the same time, Power also acknowledged that the significant capex to fund this kind of project could be challenging for a company that has had to spend the past six months selling assets and issuing equity to increase its liquidity and reduce a debt load that made Wall Street nervous. Landing these development opportunities would likely require a development joint venture with a hyperscale partner, something Power said the firm is actively pursuing.

“We are certainly around some of these opportunities ... Look at our footprint: We have extensive locations around the globe that can help those types of customers,” Power said. “Adding to that a development partner, I think, will put even more fuel on the fire to support that growth.” 

Still, some analysts are skeptical about Digital Realty’s ability to land these kind of massive AI-driven development deals. In addition to the company’s need to limit capital expenditures, TD Cowen’s July report points to Digital Realty’s relatively high pricing compared to competitors and an impediment to leasing volume. 

At Equinix, on the other hand, the firm’s leadership indicated on the company's quarterly earnings call that it is not pursuing these massive hyperscale AI leases in U.S. markets — not in the immediate future, anyway. 

Equinix’s hyperscale-focused business line — known as xScale — has exclusively focused on development in international markets, with the firm’s U.S. footprint oriented toward colocation, connectivity and other services. But while Equinix CEO Charles Meyers told analysts the company has not been actively chasing the AI-driven hyperscale opportunities that are manifesting mainly in U.S. markets, he said the potential to expand hyperscale development efforts into North America is increasingly part of the discussion. 

“I do think that our posture has probably evolved a little bit in terms of our thinking around xScale in the Americas broadly and in the U.S. specifically, and I think AI is part of that, although it's not the only factor,” Meyers said on the company's earnings call Wednesday. “We're looking at how we would do that, potentially.”

Despite not benefiting from the AI boom yet, the two major data center REITs still reported strong year-over-year growth. Equinix reported revenue of $2.02B in the second quarter, up 14% year-over-year, while Digital Realty posted $1.37B of revenue, up 24% from last year.