Photo credit: www.cnbc.com

The Data Center Market: Navigating Changes in Demand and Strategy

Data centers have increasingly become a defining feature of suburban landscapes, much like shopping malls and community sports fields. Recently, however, Microsoft’s decision to halt plans for data centers in Ohio has raised concerns about the sustainability of this burgeoning market. A report from Wells Fargo further fueled anxiety by suggesting that Amazon Web Services is reconsidering some of its proposed data center projects.

Despite these developments, it’s worth noting that the so-called ‘data center bust’ may have never truly commenced. The current shift in project timelines reflects not a downturn but rather a strategic reevaluation amid ongoing robust spending patterns.

“We are witnessing accelerated growth in AI deployments across the data center landscape, with strong indicators of demand supporting both our short- and long-term expansion,” stated Giordano Albertazzi, CEO of Vertiv, a data center supplier based in Ohio, during a recent earnings call. Following this update, Vertiv’s shares surged by 22%.

Notably, both Amazon and Nvidia reaffirmed their confidence in the strength of the data center market last week. Kevin Miller, Amazon’s vice president for global data centers, emphasized at a conference organized by the Hamm Institute for American Energy that “there has been no significant shift in our projections.” He added that demand continues to grow strongly and is expected to increase further in the upcoming years.

Nonetheless, the strategic landscape is evolving as the AI market continues to develop. Over a short span this year, several events have influenced perceptions, including the emergence of China’s DeepSeek, the announcement of President Trump’s $500 billion AI-powered Stargate initiative, and heightened concerns regarding tariffs and trade disputes.

“These developments have created a scenario where the data center sector is currently experiencing a general pause,” remarked Pat Lynch, executive managing director of CBRE’s Data Center Solutions. However, Lynch views this as a temporary hiatus, asserting that the pipeline for new projects remains significant and active, reflecting ongoing deal-making activity. He expressed optimism about future demand, especially as large AI training models gain traction.

Microsoft previously committed to a substantial $1 billion investment in Ohio’s data center landscape, an area poised for growth alongside planned Intel chip factories. However, the company has recently opted to reconsider its timeline. “After careful evaluation, we will not be advancing with our data center plans in Licking County at this time. We will continue to assess these sites in line with our broader investment strategy,” a Microsoft representative stated.

A recent UBS analysis suggested that Microsoft may have overextended itself during the AI surge and is now refining its focus on the most viable projects. The report highlighted a staggering increase in Microsoft’s leased capital expenditures—up 6.7 times over just two years—resulting in lease obligations of approximately $175 billion. “Microsoft aggressively acquired leased data center capacity throughout 2022-2024 and now has the insight to streamline some of its early-stage projects,” UBS noted, adding that there is less support for the notion of a “demand lull.”

Alphabet’s CFO, Anat Ashkenazi, described the cloud supply-demand scenario as “tight” post-earnings announcement. “We may observe fluctuations in cloud revenue growth rates correlating with quarterly capacity deployments,” she commented, estimating relatively higher deployment levels toward the end of 2025.

“What we are witnessing is not a retreat from demand but rather a strategic realignment,” explained John Carrafiell, co-CEO of BGO, a prominent global real estate investment firm with $83 billion in managed assets, particularly in data centers. Significant players like Microsoft, Google, Meta, and Amazon are on track to allocate over $300 billion in capital expenditures this year, primarily aimed at AI infrastructure, not to mention other substantial participants like OpenAI and Oracle involved in the Stargate initiative.

Carrafiell perceives the current situation as a reshuffling of priorities rather than an outright bust, noting that the critical elements of power, fiber, water, and land have become increasingly scarce and strategic. He predicts that long-term enterprise adoption will spur AI and data center demand well into the next decade, stating, “We aren’t even in the first inning yet.”

Power plays a pivotal role in data center operations, which are anything but plug-and-play. These facilities require substantial electricity for computing resources and cooling systems. As generative AI transitions from experimental phases to large-scale applications, the urgency for low-latency, high-efficiency data centers located near end-users will heighten, although the alignment of favorable conditions necessary for expanding data center footprints may take time.

“New data centers are now so large that existing power grids cannot keep pace,” remarked Allan Schurr, chief commercial officer of microgrid developer Enchanted Rock. Just three years ago, a sizable data center might demand around 60 megawatts—sufficient to power about 20,000 homes—whereas today, new facilities are seeking at least 500 megawatts.

This escalation in electricity requirements coincides with rising demand from manufacturing and the broader electrification of transportation, posing challenges for utility providers tasked with ensuring power availability during peak usage. “This explains why some utilities are reporting extended interconnection wait times for data centers,” Schurr indicated. “Utilities are now compelled to enhance substations and may need to invest in upgrading transmission and generation capabilities, a process that is time-intensive.”

CBRE’s data reveals a significant shift, with data centers expanding from just 2% of its portfolio in 2022 to an anticipated 10% in 2024. Lynch expects this trend to continue, emphasizing that the accessibility of ample power resources is critical in the current market as data center developers seek locations with robust energy availability. Areas like Georgia, Texas, and Ohio meet many of these requirements, and regions lacking the necessary grid capacity will need to scale quickly to attract investment.

“Having substantial power availability within 36 months is highly appealing to potential clients,” Lynch noted.

Approximately 3% of the global power supply is currently allocated to data centers, according to Datacenters.com.

Schurr pointed out that Enchanted Rock’s data indicates there is, by and large, sufficient power to meet this demand—most of the time. “Of the 8,760 hours in a year, the grid faces stress only during a limited fraction,” he explained. “By alleviating demand during those 100 to 500 hours, we can significantly reduce long interconnection delays.”

According to Pankaj Sachdeva, a senior partner at McKinsey & Company specializing in data center development, it is crucial to differentiate between the potential for a larger slowdown and the recent pauses adopted by major tech firms. Based on McKinsey’s recent models—omitting potential tariff impacts—the data center market is projected to grow between 20% and 25% over the next five to seven years. However, growth rates may not exhibit linear progression year over year.

Adjustments to tariffs could exert new pressures on AI and data center supply networks, especially with anticipated tariffs on critical minerals looming. “These shifts will heighten hardware costs, affect sourcing strategies, and compel companies to reevaluate their long-term procurement approaches,” added John Archer, a senior delivery principal and supply chain transformation lead at Slalom Consulting. In the near term, AI and cloud services providers will likely need to implement cost-mitigation strategies, including renegotiating supplier contracts and optimizing inventory management.

“In the long run, we can expect a movement toward geographical diversification, co-manufacturing in tariff-friendly areas, and deeper integration of AI-driven supply chain analytics to adapt to evolving trade policies,” Archer concluded.

Despite some specific project setbacks, the overarching demand for compute power—vital for AI software and hardware—remains high. Suresh Venkatesan, CEO of POET Technologies, a company dedicated to developing power solutions for data centers, stressed the necessity for more efficient solutions. “The surge in AI requires an unprecedented volume of compute power,” he stated. “While some data center initiatives may face delays, others are likely to emerge, as there is no sign of a decrease in demand for connectivity.”

Source
www.cnbc.com