Mediaboss Marketing

The rapid adoption of new artificial intelligence apps and an intensifying bid for dominance among tech giants Amazon, Google and Microsoft are expected to drive investment and double-digit growth for the data center industry in the next five years.

A gold rush of AI these days centers on the brisk development of tools such as ChatGPT, according to a new analysis from real estate services firm JLL. Voice- and text-generating AI apps could transform the speed and accuracy of customer service interactions and accelerate demand for computing power, as well as the systems and networks connecting users that data centers provide, the real estate firm said.

The emergence of AI comes on the heels of increased usage of data centers in the past few years, as people spend more time online for work and entertainment, fueling the need for these digital information hubs, which provide the speed, memory and power to support those connections.

JLL projected that half of all data centers will be used to support AI programs by 2025. The new AI applications need for enormous amounts of data capacity will require more power and expanded space for the data center services, particularly colocation facilities, which are a type of data center that rents capacity to third-party companies and may service dozens of them at one time. It's also a potential growth area for commercial property investors.

We expect AI applications, and the machine learning processes that enable them, will drive significant demand for colocation capabilities like those we provide, Raul Martynek, CEO of Dallas-based DataBank, told CoStar News in an email. Specifically, the demand will be for high-density colocation and data centers that provide significantly greater concentrations of power and cooling.

One kilowatt hour of energy can power a 100-watt light bulb for 10 hours, and traditional data server workloads might require 15 kilowatts per typical cabinet, or server rack, Martynek said. But the high-performance computing nodes required to train large language models like ChatGPT can consume 80 kilowatts or more per cabinet.

This requires more spacing between cabinets to maintain cooling, or specialized water-chilled doors to cool the cabinets, Martynek said.

In addition to the added energy and water needs, the growth in data centers faces other challenges. Credit-rating firm S&P Global Ratings noted that long-term industry risks include shifting technology, cloud service providers filling their own data center needs, and weaker pricing. The data center industry, with power-hungry facilities running 24 hours a day and 365 days a year, has also received criticism from environmentalists.

DataBank owns and operates more than 65 data centers in 27 metropolitan markets. This month, it secured $350 million in financing from TD Bank to fund its ongoing expansion.

It was DataBanks second successful financing this year, coming just weeks after completing a $715 million net-lease securitization in March 1. Under net-lease offerings, issuers securitize their rent revenue streams into bonds. The sale of those bonds replenishes the issuers capital to be used to pay down debt and continue investments.

ChatGPT and other apps are bots that use machine learning to mimic human speech and writing. ChatGPT debuted in November and is most arguably the most sophisticated to launch so far. AI software developer Tidio estimated recently that usage of such bots has already grown to 1.5 billion users worldwide.

In January, Microsoft announced a new multibillion-dollar investment in ChatGPT maker OpenAI. Google has recently improved its AI chatbot, Bard, in an effort to rival its competitors. And Amazon Web Services, the largest cloud computing provider, introduced a service last week called Bedrock aimed at helping other companies develop their own chatbots.

Amazon CEO Andy Jassy touted the e-commerce giants AI plans in his annual letter to shareholders.

Most companies want to use these large language models but the really good ones take billions of dollars to train and many years and most companies dont want to go through that, Jassy said last week on CNBC. So what they want to do is they want to work off of a foundational model thats big and great already and then have the ability to customize it for their own purposes. And thats what Bedrock is.

The growth projections of AI have data center owners and operators at the forefront of the securitized bond market. Three data center providers have issued $1.3 billion in net-lease securitized offerings already this year, according to CoStar data. Thats more than all of last year combined. In addition, two more providers have offerings in the wings.

The sector is a bright spot in an otherwise weakened market for other commercial real estate securitized bond offerings, down more than 70% from the same time last year.

The data center space remains extremely attractive to capital sources looking for quality and stability versus other asset classes that have been challenged amidst uncertain economic conditions, Carl Beardsley, managing director and data centers lead at JLL Capital Markets, told CoStar News in an email.

JLL said data center financing comes from a variety of sources including debt funds, life insurance companies, banks and originators of commercial-mortgage backed securities.

Although money center banks and some regional banks have become more conservative during this volatile interest rate period, there is still a large appetite from the lender universe to allocate funds toward data centers, Beardsley said.

JLL is forecasting that the global data center market is expected to grow 11.3% from 2021 through 2026.

Across its six primary data center markets Chicago, Dallas-Fort Worth, New Jersey, Northern California, Northern Virginia and Phoenix the United States has a strong appetite for data centers property transactions compared to other countries, according to JLL, accounting for 52% of all deals from 2018 to 2022. These markets also have a data center capacity of 1,939 megawatts under construction, JLL said. One megawatt is equal to 1,000 kilowatts.

The growth is expected to continue even heading into a potential recession, according to S&P, which has rated two of the three data center securitized bond offerings completed this year so far.

Overall supply and demand is relatively balanced as new data center development has been constrained in certain markets by site availability, lingering supply chain issues and more recently, power capacity constraints, S&P noted in its reviews. Although we expect data centers to see some growth deceleration in a recessionary environment, we believe it will be mitigated by the critical nature of data centers.

S&P added that market data suggests 2022 vacancy rates were low for key data center markets and rental rates increased year over year.

New net-lease securitized fundraisings this year have come from DataBank, Stack Infrastructure, and Vantage Data Centers.

Denver-based Vantage, a global provider of hyperscale data center campuses, saw unprecedented growth in 2022, outperforming its previous record set in 2021. The company began developing four new campuses internationally and opened 13 data centers. The company raised more than $3 billion last year to support that effort.

Last month, Vantage completed an additional securitized notes offering raising $370 million. The offering was backed by tenant lease payments on 13 completed and operating wholesale data centers located in California, Washington state and Canada.

Stack, a Denver-based developer and operator of data centers, issued $250 million in securitized notes last month.

Stacks growth is outpacing the industry with a portfolio of more than 1 gigawatt, or 1,000 megawatts, of built and under-development capacity, and more than 2 gigawatts of future development capacity planned across the globe. The company has more than 4 million square feet currently under development.

Stack most recently announced the expansion of a Northern Virginia campus to 250 megawatts, the groundbreaking for another 100 megawatt campus in Northern Virginias Prince William County and the expansion of its 200 megawatt flagship Portland, Oregon, campus.

In addition, Dallas firm CyrusOne and Seattle-based Sabey Data Centers have filed preliminary notices of offerings in the works with the Securities and Exchange Commission.

See the original post:
'Gold Rush' in Artificial Intelligence Expected To Drive Data Center ... - CoStar Group