Powering the AI Revolution: How Energy Availability Is Shaping the Next Phase of Data Center Expansion

This edition of the Investment Stewardship in the AI Era series addresses the question of whether the infrastructure required to support AI-driven demand can be delivered and how infrastructure constraints may shape growth. The report uses asset-level data from the ISS STOXX Geospatial Database to examine where data center infrastructure is concentrated, how operators are exposed to different electricity systems, and what this may mean for future growth. 

By moving beyond aggregate demand projections, Powering the AI Revolution highlights how local factors such as power availability, grid capacity, and permitting may increasingly shape the development of AI infrastructure. For investors, the analysis offers a perspective that is often absent from company reporting, helping to identify where infrastructure constraints may affect the delivery of future capacity.

Key Takeaways

• AI infrastructure is becoming constrained by power, not capital. Investment in data centers continues to accelerate, but future expansion increasingly depends on access to power, grid connections, and supporting infrastructure. The ability to secure electricity is becoming an important determinant of where growth can occur. 
• Announced capacity should not be treated as delivered capacity. The value of a development pipeline ultimately depends on whether projects can secure power and connect to the grid. Connection queues, permitting requirements, and infrastructure constraints can delay when assets become operational and start creating value for investors. Increasingly, the difference between operators lies less in the size of their pipelines than in the electricity systems on which those pipelines depend.
• Exposure is local, and similar growth plans can face different constraints. Data centers are a small share of global electricity consumption but a large share within the hubs that host them. Operators with comparable country footprints may face very different conditions depending on which electricity systems their assets depend on. For investors, similar geographic footprints do not necessarily translate into similar operating conditions, as local power systems can differ significantly.
• The ability to convert demand into delivered capacity may vary across business models. Hyperscalers (large cloud and AI platform providers), co-location providers (companies that own and lease data center capacity), and crypto-focused operators approach growth differently. Their ability to secure capacity, adapt to electricity constraints, and manage development timelines can materially influence future outcomes.
• Asset-level analysis provides insight beyond company disclosures. Many of the factors that determine future capacity delivery, including connection status, queue position, and access conditions, remain largely outside public disclosures.