By Joshua Burd
Demand for data center space in New Jersey is surging on the back of Big Tech and the artificial intelligence arms race. And while a lack of electrical capacity may keep a lid on new supply, the owners of existing facilities are now racing to expand their footprints.
It’s also a new opportunity for those with developable land — at least the kind that can support the complex, highly specialized needs of AI platforms and other innovations.
“We’ve been trying to get as smart about this segment as we can because of the number of inquiries that we’ve been seeing, on sites that we own or are under contract to buy, for this particular use,” said David Greek, a managing partner of East Brunswick-based Greek Real Estate Partners. That interest has highlighted the “specific items” that data center users are seeking — namely, access to power and security — and “there is such a limited set of sites that have that in place already, that the users know exactly where they are.”
Data center vacancy in the New York tristate area fell to an all-time low in the second half of 2023, to 6.5 percent, according to a report by CBRE, thanks largely to demand tied to artificial intelligence. The firm noted that AI companies preleased more than 40 megawatts in the region during that period, with notable deals including a lease at QTS’ expanded facility in Piscataway and the preleasing of a 22-megawatt, 554,000-square-foot project in East Windsor.
Meantime, rental rates in the region rose during the second half of 2023 to $130 to $150 per kilowatt for a three- to 10-megawatt requirement per month, CBRE found.
“We’re starting to see a real supply shortage,” said Jon Meisel, a senior vice president with CBRE. He noted that many of New Jersey’s longstanding data center owners have shown restraint in recent years with respect to new development, following a building boom more than a decade ago that was fueled by big banks, media and other sectors. Users have been “nibbling away at our inventory” ever since, as other U.S. markets have blossomed with the help of lower utility costs, government incentives and large-scale requirements from tech firms.
“These folks who were very judicious have run out of capacity,” said Meisel, whose team was marketing a fully approved, 21-acre site in Totowa that can support a 40-megawatt data center across 300,000 square feet.
Developers are now expanding where they can. JLL broker Tom Reilly noted that roughly 70 megawatts of capacity is coming to New Jersey this year by way of new construction, including a new 300,000-square-foot project at Equinix’s campus in Secaucus, as well as CoreSite’s newly approved, 85,000-square-foot facility on Emerson Lane in the town that will add another 15 megawatts.
Meantime, QTS’ new preleased facility in East Windsor, which is redeveloping a former McGraw Hill Financial complex on Princeton-Hightstown Road, follows a deal that filled the remaining space at its expanded complex on Possumtown Road in Piscataway. Brokers also point to a new Databank facility just outside New Jersey, in Orangeburg, New York, which was preleased to an undisclosed AI company and will give way to a 30-megawatt second phase.
“The good news is it’s going to add significant inventory,” Reilly, a managing director with JLL, said of the state’s development pipeline. “The bad news is, for us, a lot of it is pre-committed.”
He also pointed to an abundance of projects that are in planning stages but have yet to break ground, which would add almost 150 megawatts of inventory. According to a recent report by JLL, those projects equate to around 850,000 square feet of new development, which would create coveted supply for AI deployments as well as life sciences and research users.
Meeting that demand is inherently complicated in New Jersey, with its high utility costs and a power grid that is already stretched. The challenge is only more acute with users seeking more scale than ever before to support AI platforms, such as ChatGPT and other large language models, requiring higher-density data centers that consume two or three times as much power per square foot.
Essentially, developers that were content to build 40- to 50-megawatt projects in past cycles may now need to think bigger — if the grid supports it. All the while, speed to market is still critical.
“What happens is that, because the supply has dried up so quickly, there’s not a lot of areas that have a confluence of ample power and approvals or a box ready to go for AI,” Meisel said. “And what the AI folks are chasing right now is a clear, risk-free path to becoming one of the first on the ground.”
That hasn’t stopped developers from trying to meet the market. Greek Real Estate Partners, for instance, is increasingly conscious of which sites in its portfolio could accommodate a major data center, given the interest it has fielded in recent months. The developer is now actively engaging those tenants, which range from government agencies and tech companies to large financial institutions.
“It’s become a bigger part of our marketing efforts now when we recognize the sites that have that infrastructure and seeing the amount of demand that sector’s throwing off,” GREP’s Greek said. “That has become central to some of our leasing strategy, especially in the preconstruction phase of a project where we might be finishing entitlements or we might be doing work on the site that isn’t actually building the building.”
The firm has “no immediate plans” to build data centers speculatively, he added, but even a build-to-suit can require three or four years of lead time. Users are starting their search increasingly early because “they’re so incredibly complex to build,” while it can be a lengthy process to secure commitments from a utility company.
Langan, the engineering and environmental services firm, also notes that a data center’s cooling systems often require a large amount of water.
“These demands and existing utility infrastructure are the first thing a data center developer will look at when assessing a site,” said Dan Miola, a principal with Langan, which performs a host of services to vet the feasibility of the site. That includes early discussions with utility companies, environmental and geotechnical investigations and meetings with the local agencies.
“This vetting helps our clients identify the main project challenges related to feasibility, cost premiums and schedule,” Miola added.
Naturally, Greek said, new data centers would be competing with traditional warehouses and distribution centers, whose own electrical needs have grown in recent years with the rise of automation. That means there’s likely to be “some winners and losers when it comes to power supply over the next 10 years, and it’s going to require a lot of work and a lot of investment from the utilities just to keep pace with the amount of demand in the market.”
As for why it’s worth the trouble to try to build or lease a data center in New Jersey — especially compared to other markets — Reilly noted that “there are going to be certain applications that need to be close.” That includes AI technologies that are specific to oncology and other health care uses, he said, as well as engineering research. That’s not to mention “the next generation of what we’re doing on our devices” and even the prospect of self-driving vehicles.
“That will drive proximity … all of the things that will be tied to AI and what we do that’s got to be local,” Reilly said.
Greek also pointed to New Jersey’s skilled labor force, its top-tier academic institutions and its proximity to corporate headquarters. All of which figure to attract a broad range of data center users, especially as new industries look to leverage AI and develop tools that are specific to their needs.
“It will continue to evolve,” Greek said. “I think we’re still in relatively early innings of how these facilities work, how much they cost to build and how many of them we need. It is hard to predict what the demand will be based on the fact that it’s being driven by a technology that, frankly, nobody really knows what its larger impact on the economy will be.
“Some people are saying it’s a game changer,” he added. “And if that is really true for productivity, then I think we’re going to be building a lot of these over the next 10 years.”
