Demand for artificial intelligence (AI) and cloud data centers surged in the first half of 2024, in both established and emerging markets. Absorption is poised to surpass the record levels set in 2023. Power availability remains a critical concern, prompting developers to explore extensive geographic areas for substantial power options within the next two to three years. Despite an expanding supply pipeline, demand continues to outpace supply, resulting in consistently declining vacancy rates across the board.
“Interest in large-scale power availabilities, plentiful land and less strict latency requirements for AI, has driven hyperscalers and operators to expand in a host of historically peripheral markets such as rural Georgia, North Carolina, Pennsylvania, Texas, Minnesota and the Dakotas, among many other outlying areas just beyond the ‘major’ established data center markets,” said Bo Bond, Cushman & Wakefield’s Executive Managing Director. “Active areas of the country are being defined more by the utility provider service area than a more traditional U.S. metropolitan development marketplace.”
Where utility providers have been unable to provide power sooner, certain operators have collaborated with power companies to deliver substations, transmission lines or source micro-grid power. Many of these agreements are now being signed directly with third-party energy generation developers, with wind, solar, battery storage, natural gas and even geothermal developments moving quickly across markets. Going forward, Cushman & Wakefield expects to see more operators continue to acquire large acreage high-power capacity sites in even more rural markets. Additionally, some operators will begin to work to secure power along longer timelines.
“AI data centers are increasingly becoming integral to both hyperscale and colocation development pipelines. Although hyperscalers remain the primary end users in this space, several GPU cloud providers are entering the market, seeking hyperscale-level capacity across the Americas,” said Ali Greenwood, Cushman & Wakefield’s Executive Director.
AI training facility plans primarily focus on large, latency-tolerant locations in rural markets, while AI inference facilities are strategically positioned near major cloud regions. As hyperscale growth accelerates in both established and emerging markets, expect further developments in newly emerging data center markets. This trend will be driven by providers and investors prioritizing power transmission and hyperscaler deployment.
At an overall stock level, Virginia remains far and away the largest data center market in the world, with 13.2GW (combined existing, under construction and committed pipeline), followed by the other primary markets in the space: Atlanta (3.8GW), Phoenix (2.9GW), Chicago (2.4GW), Dallas (2.3GW) and Silicon Valley (1.3GW), which all have seen continued expansion despite varying levels of limitations to power availability.
Portland and Eastern Oregon (1.9GW), and Columbus, Ohio (1.8GW), have quickly risen as comparably large markets driven by both hyperscale self-build and colocation deployments.
