Soil over silicon: Why data center forecasting is a real estate problem

For years, the industry has been obsessed with "tech" metrics—FLOPs, GPU clusters, and the bottomless appetite of Large Language Models. But as we move into 2026, the bottleneck has shifted from silicon to the soil beneath the concrete slab.

While tech metrics dominate the headlines, the real challenge for energy companies and grid operators isn't just estimating how much power the industry wants; it’s determining how much power the physical landscape can actually support.

The phantom gigawatt problem

The gap between ambition and reality is widening. On paper, the market is tracking roughly 190GW across 777 large-scale data centers and AI factories announced since 2024. Of that, 16GW is slated to hit the grid in 2026.¹

However, the "boots on the ground" reality reveals a sobering truth:

• Only 5GW of that 16GW is currently under construction.²

• The remaining 11GW exists only as speculative press releases with no visible construction progress.³

• Given typical build timelines of 12 to 18 months, any project not currently breaking ground is at high risk of missing a 2026 delivery date.⁴

• Nearly half of the capacity slated for 2026 remains in the "announced" stage, signaling massive upcoming delays.⁵

For utilities trying to manage grid stability, these "phantom gigawatts" are a systemic risk to infrastructure planning. In 2025 alone, 26% of expected capacity slipped and 10% of projects delayed their operation dates without warning.⁶

Trading press releases for property data

To avoid planning for mirages, energy companies must stop treating data center forecasting as a software projection and start treating it as a real estate and permitting challenge. Reliable forecasting requires tracking physical signals rather than just megawatt requests.

By integrating Cotality’s high-fidelity property data, utilities can finally see the truth of a project’s viability through four key pillars:

• Building permits and zoning changes: A press release takes an afternoon to write, but a zoning change can take years. If a gigascale campus is announced but the land is still zoned for agricultural use, that load is likely years away.

• Land acquisition: Tracking ownership changes and parcel subdivisions allows energy companies to see when a hyperscaler has actually closed on a property. When a major tech firm begins consolidating contiguous parcels, it’s a high-probability signal that real load is coming.

• Infrastructure proximity: Data centers require specific proximity to fiber backbones and high-voltage transmission lines. Cotality allows utilities to assess the physical feasibility of a site; if the distance to a substation is too great, the project will likely face interconnection bottlenecks.

• Hazard assessment: Environmental risks like flood zones, seismic activity, and local community opposition can kill a project in the eleventh hour. Property-level data helps energy companies discount the probability of these projects ever reaching the "turn-on" phase.

Moving beyond the headlines

The era of taking data center announcements at face value is over. When 30% to 50% of the pipeline is unlikely to materialize on time, the most valuable tool in an energy company’s kit isn’t a speculative AI forecast—it’s a map.⁷

By integrating Cotality’s granular data on permits, land transitions, and infrastructure proximity, energy providers can stop reacting to headlines and start planning around real-world construction.