AI Lives in Data Centers
Power Constraints and AI Computing Reshape Data Center Industry
The data center real estate sector stands at a critical juncture where traditional growth models are being disrupted by power availability constraints and AI computing demands. The data shows storage capacity in data centers will grow from 10.1 zettabytes in 2023 to 21.0 zettabytes by 2027, representing an 18.5% compound annual growth rate. This growth trajectory creates significant real estate implications.
Power availability has become the primary constraint on new development. In key markets like Ireland, data centers now consume 18% of national electricity, leading to connection moratoriums. Similar constraints are evident in Singapore and Northern Virginia, where power demand from data centers is expected to double by 2028.
The economics of data center real estate are being fundamentally altered by AI computing requirements. Average rack density in hyperscale facilities is currently 36kW per rack and projected to approach 50kW by 2027. This represents a material change in building specifications and infrastructure requirements. New AI workloads are pushing some installations to 80-100kW per rack, necessitating liquid cooling systems and redesigned power distribution.
From an investment perspective, these trends create both opportunities and risks. While demand remains robust with Structure Research projecting 15.2% CAGR in global colocation capacity, power constraints are extending development timelines and increasing costs. Markets with abundant power resources and renewable energy access are seeing increased activity as operators seek viable development sites.
The sustainability imperative adds another layer of complexity. With data centers accounting for 1% of energy-related greenhouse gas emissions, operators must balance reliability with environmental impact. This is driving innovation in facility design and cooling systems, with liquid cooling showing potential 90% power reductions for cooling operations.
For real estate investors, the key implications are:
Site selection now primarily driven by power availability rather than traditional real estate metrics
Higher development costs due to enhanced power and cooling infrastructure requirements
Increased focus on secondary markets with available power capacity
Premium values for facilities that can accommodate high-density AI computing
Growing importance of sustainable power access in asset valuation
The data center sector remains an attractive real estate investment category given strong demand fundamentals. However, success will increasingly depend on solving power availability challenges while meeting enhanced technical specifications for AI computing. This suggests a more specialized approach to development and investment than historical patterns.
