North American Data Center Vacancy Hits Record 1.4% as AI Demand Absorbs New Capacity

According to CBRE's North America Data Center Trends H1 2026 report, primary-market data center supply surged 33.7% year-over-year to reach a record 10,903 MW, yet vacancy dropped from 1.6% to an unprecedented low of 1.4%. Net absorption climbed 11.7% to 1,456.2 MW as hyperscale cloud providers and AI infrastructure developers absorbed capacity as fast as it was energized. Total under-construction capacity expanded 24.8% to 7,481.1 MW, with Atlanta overtaking Northern Virginia as the top market by under-construction volume (2,882 MW). Crucially, 80.4% of all under-construction capacity across primary markets is already preleased, leaving less than 1,500 MW of future supply uncommitted. This supply bottleneck fundamentally changes how platform engineers and infrastructure leads procure compute and colocation footprint. Near-term large contiguous power blocks (5 MW and above) are virtually unavailable across mature hubs like Northern Virginia, Silicon Valley, and Chicago through late 2027. The severe shortage has driven substantial rental rate inflation down the stack: asking rates for smaller 250-to-500 kW deployments rose 4.3% in H1, while mid-range 3-to-10 MW requirements jumped 8.3%. High compute density and aggressive AI model training/inference requirements are no longer just squeezing specialized AI labs; they are exerting heavy pricing and availability pressure on conventional enterprise workloads. This development highlights the broader structural transformation of data center infrastructure from commoditized commercial real estate into power-constrained, grid-dependent utility assets. As traditional tier-one metros encounter local electrical substation bottlenecks, permitting pushbacks, and land scarcity, data center operators are shifting focus toward emerging frontier regions such as the Southeast, Indiana, and Texas. Furthermore, developers are increasingly adopting modular delivery workflows and "bring your own power" (BYOP) generation models to compress deployment schedules that now span multiple years for massive campus-scale builds. In practice, engineering organizations must move away from just-in-time infrastructure expansion. DevOps and site reliability engineering (SRE) teams should prioritize workload efficiency, model quantization, and bin-packing optimization to maximize existing capacity allocations. Platform architects must design systems for multi-region and multi-provider topology, leveraging secondary markets and edge points of presence rather than waiting for contiguous blocks in flagship data centers. Finally, procurement teams must prepare for restrictive lease structures—including strict take-or-pay power floors and extended multi-year forward commitments—when planning future footprint expansions.