Hyperscale data centers scale up grid, power needs as AI accelerates

Solmar Insights

The global hyperscale data center market is rapidly expanding to meet the needs of next-generation computing, with a compound annual growth rate of 6.2%. Facility counts worldwide grew from 1,136 at the end of 2024 to 1,297 by the third quarter of 2025, and power requirements for AI workloads are driving even larger increases in total capacity and grid demand.

Facility growth and shifts in scale

While data center facility counts continue their climb, the most notable shift is the dramatic surge in their scale and capacity. From the end of 2024 to Q3 2025, the number of global hyperscale sites grew by 161, with more than 500 additional projects planned or underway. However, the expansion of compute and power capabilities is outpacing just the facility count: new AI-driven data center campuses are being planned for hundreds of megawatts, with some announcements surpassing the 1,000 MW milestone for a single development.

This push toward mega-scale facilities means that large campuses can effectively replace many traditional, smaller data centers, while simultaneously introducing more stringent requirements for grid connection, energy infrastructure, and build timelines. Phased construction is increasingly used to accelerate delivery of usable capacity ahead of full site completion.

Rising power demand for AI

The rise of artificial intelligence workloads is fundamentally reshaping the requirements and design standards for hyperscale data centers. AI model training and inference depend on clusters of accelerators connected via high-bandwidth, low-latency networks, driving significant increases in electricity demand per rack and across entire campuses. In 2026, leading AI data centers collectively deliver 11.4 GW in IT power and operate over 12.1 million advanced H100 accelerators.

Rack power consumption, traditionally less than 10 kW, now routinely exceeds 50 kW in intensive AI deployments, necessitating not only higher electrical supply but radically different architectures for cooling and network infrastructure. Some data center operators are actively designing facilities on the scale of 1,000 MW, given the compounding requirements for both compute density and power redundancy.

Grid connectivity as a market constraint

A critical new constraint is the availability of grid power and transmission: as planned hyperscale capacity is set to rise more than sixfold from 33.96 GW in 2025 to 228.96 GW by 2035, electrical infrastructure is quickly becoming the leading driver for site selection in key markets. Large power interconnection requests require multi-year planning, direct coordination with utilities, and, in many instances, upgrades to local transmission networks.

Investors and developers are increasingly prioritizing regions with robust grid accessibility and streamlined interconnection processes. In turn, power grid capacity and resilience now directly impact the feasibility and speed to market of new hyperscale projects. The shortage of available transmission in some North American markets is shifting attention to areas with surplus power, as operators seek to avoid protracted delays for approval and energization.

Cooling and network challenges at scale

The increase in compute density is also transforming approaches to thermal management and facility layout. Traditional air cooling methods are facing limits in their ability to handle high-density deployments; as a result, state-of-the-art hyperscale data centers are deploying liquid cooling systems to efficiently dissipate heat and maintain optimal conditions for dense AI clusters.

Higher compute density also dictates new requirements for intra-campus networking, including the deployment of ultra-high-speed, low-latency connections between racks and clusters. These technical shifts impact layout and utility infrastructure, influencing everything from water consumption profiles to the complexity and reliability of facility operations.

Outlook through 2035 for investors

The forecasted leap in hyperscale capacity by 2035 underscores the competitive importance of aligning digital infrastructure strategies with local grid realities. With more than 500 projects in pipeline globally, development risks now hinge as much on transmission buildout and permitting as on traditional land or equipment constraints. The expanding prominence of power procurement and grid interconnection contracts highlights a shift in competitive advantage from facility construction alone to energy and connectivity partnerships.

Institutions with access to reliable grid resources are positioned to advance projects at scale, while those facing congested or uncertain grid environments will encounter delays and higher risk profiles. For buyers, timely due diligence on utility availability and support for innovative on-campus energy solutions is now essential to achieving capacity objectives in prime hyperscale markets.

Reporting via the original publisher