Capital continues to surge into AI data center development at a pace rarely seen in digital infrastructure. Developers and hyperscalers announce multi gigawatt campuses every week. These announcements create momentum, but they often mask a large data center capacity gap between headline numbers and real, committed capacity.
This post examines why the data center capacity gap persists, how phantom demand shapes utility planning, and what this means for developers, suppliers, and investors.
Why the Data Center Capacity Gap Exists
Developers often announce projects early. They secure land, identify potential anchor tenants, and outline a high level power strategy. These steps help them build visibility and attract interest. However, construction starts only after they secure critical path items. These include power generation equipment, transformers, switchgear, municipal permits, and a committed anchor tenant.
The development lifecycle typically moves through four stages:
- Land acquisition and early strategy work across customer, power, and financing.
- Equipment orders for turbines, transformers, and switchgear with multi year lead times.
- Construction once permits, financing, and delivery schedules align.
- Operations after the initial phase is complete.
A headline announcing a 1 GW campus usually reflects only 100 – 200 MW in active construction. Industry data shows that roughly 10% of committed capacity sits under construction at any given time. The share of announced capacity under construction is even smaller, which widens the data center capacity gap.
The scale of the gap is visible in grid data. In November 2025, ERCOT tracked 226 GW of large load interconnection requests, up from 63 GW just one year earlier. To put that in perspective, the historical peak electricity demand in Texas is 85.5 GW. This means the current data center queue is 2.6 times larger than the maximum power the state has ever needed. Yet, of that massive 226 GW pipeline, only 7.5 GW is actually connected and operating today. That represents a realization rate of just 3.3%, proving that the vast majority of announced capacity remains purely theoretical.
Phantom Demand and Its Role in the Capacity Gap
The development timeline explains part of the data center capacity gap. The rest comes from phantom demand. Utilities such as American Electric Power (AEP) and Dominion Energy have described this dynamic publicly.
Developers now face supply chain constraints rather than capital constraints. They evaluate several sites for a single project. Instead of waiting to choose one location, they file interconnection requests with multiple utilities across multiple states. They negotiate and eventually select one site. The other requests often remain in the queue without formal withdrawal.
These unused requests appear as live megawatts. They distort the picture of future demand for every utility and grid operator. As a result, interconnection queues contain large volumes of speculative load that may never materialize.
Examples That Reveal the Capacity Gap
Several high profile projects highlight the difference between announced capacity and committed reality:
- xAI Colossus in Memphis: The project launched with-multi gigawatt ambitions. Portions operate today, but expansion follows a phased approach. Temporary mobile gas turbines support early operations while permanent infrastructure arrives. Equipment availability sets the pace.
- Microsoft Pecos Campus (Texas): Announced as a ~2 GW data center campus to support AI workloads, representing one of Microsoft’s largest single capacity additions. Development is phased over five to seven years, with power supplied via a co-located natural gas project (in partnership with Chevron) expected to deliver first power in 2028. Full announced capacity will roll out incrementally tied to infrastructure and equipment availability.
- Galaxy Digital/CoreWeave Expansions (Helios Campus, Texas): Announced multi-GW scale ambitions across the U.S. with contracted power capacity exceeding 3 GW. Individual campuses, such as Helios, proceed in phases: Phase I delivering 133 MW critical IT load in 2026, with subsequent phases ramping up in 2027 and beyond. This reflects the typical pattern of staged buildouts dependent on power delivery and construction timelines.
- Google Texas Projects (Meitner Energy Center / Panhandle campuses): Part of a $40 billion commitment through 2027, including multiple large data center campuses (often 500 MW – 1+ GW scale) co-located with new energy generation (wind, solar, battery, and gas). Projects advance in phases alongside dedicated power infrastructure, with full buildout extending years beyond initial announcements.
- Meta Alberta: Announced as up to a 1.8 GW campus with an initial 1 GW-scale phase and a new dedicated natural gas generation facility (Pembina Greenlight). The permanent power plant is not expected until around 2030. Headlines focus on the ultimate capacity, while actual construction and initial operations begin with a smaller grid-supported phase before scaling.
The Equipment Bottleneck and Its Impact on the Capacity Gap
Capital is abundant. Equipment is not. Turbines, transformers, and switchgear carry long lead times. These constraints create the true bottleneck in the market and widen the data center capacity gap.
Developers who place firm, non-cancellable orders for turbines or transformers show real commitment. Developers who do not place these orders signal that their announced capacity remains theoretical. This reality reshapes how investors evaluate project pipelines and how suppliers forecast demand.
How Utilities Respond to the Capacity Gap
Phantom loads force utilities to rethink how they manage interconnection queues. If utilities build infrastructure to serve speculative requests, they risk stranded assets and higher costs for ratepayers.
Utilities now push for stricter requirements. In September 2025, AEP’s Ohio Power Company updated its pipeline disclosure. Interconnection requests dropped from more than 30 GW to 13 GW. Many removed requests failed to meet new criteria.
Utilities also increase deposits. ComEd in Illinois now charges up to $1 million to secure a queue position for 50 MW or more. Regulatory pressure rises as well. Texas Governor Greg Abbott directed the Public Utility Commission to ensure data centers pay for the grid infrastructure they require.
Concluding Thoughts: Implications for Investors
The data center capacity gap creates analytical challenges for investors in digital infrastructure and power.
Several implications stand out:
- Announcements do not equal commitments. Equipment orders, transformer lead times, and firm deposits provide a clearer view of future supply than press releases.
- Phantom demand and the data center capacity gap make it harder to estimate real AI power supply, which complicates any forecast of future compute capacity. Accurate modeling now requires deeper insight into interconnection progress, equipment procurement, and onsite generation plans. [See here for insights on forecasting data center power demand.]
- Compute efficiency changes the supply picture. GPUs and custom ASICs deliver more performance per watt each generation. Compute output rises even if physical power supply grows slowly. AI capacity may meet market needs despite delays in data center power buildouts.
- Regulatory response is accelerating and reshaping queue economics in real time. AEP Ohio’s pipeline cut, ComEd’s higher deposits, and Governor Abbott’s directive to the Texas PUC show utilities and regulators actively filtering speculative requests. A project’s regulatory standing is now as relevant to diligence as its technical readiness.
- Investors should distinguish site developers from anchor tenants. Some developer pipelines are actually built by separate infrastructure owners who lease capacity to the named tech company. These structures carry different risk profiles and should not be evaluated the same way.
- Investors should benchmark peers on comparable data. A developer citing gigawatts under development is not directly comparable to one reporting fully contracted, interconnection‑cleared capacity. Investors should normalize disclosures before comparing scale or execution across companies.
These realities make power due diligence essential. Investors need to verify the credibility of a project’s power path by confirming interconnection progress, equipment procurement, onsite generation plans, and the timeline for transformers and switchgear. Strong power due diligence separates real projects from optionality and shows which developments can actually deliver compute on schedule.