Data center under construction

The 7 Gigawatt Gap Becomes the Defining Metric of the Cycle

Grid Stress, Storms and Resilience Economics
May 06, 2026•7 min read

Sightline Climate's pipeline tracking shows roughly 30 to 50 percent of announced 2026 data center capacity will not energize on schedule. The shortfall is not a market story. It is a real estate underwriting story.

By Keith Reynolds | Publisher & Editor, ChargedUp!

Home | All Stories

As of this past April, the gap between announced AI data center capacity and capacity actually under construction has widened to roughly seven gigawatts. Sightline Climate's project database, which tracks 190 GW across 777 large data centers and AI factories announced since January 2024, finds that of the 12 to 16 gigawatts targeted for energization in 2026, only about 5 GW is under active construction. Industry analysts now expect 30 to 50 percent of the announced 2026 pipeline to slip into 2027 or beyond. The actual energization rate by year-end may close to 20 percent rather than 33 percent.

The shortfall has become the defining metric of the cycle. At typical hyperscale densities of 100 to 300 megawatts per campus, the missing capacity is equivalent to roughly 30 to 70 large AI training facilities that were promised but cannot be delivered on schedule. Each campus represents $1 billion to $4 billion in capital expenditure. The deferred construction value runs into the tens of billions of dollars for 2026 alone. Olivia Wang, research analyst at Sightline Climate and co-author of the Outlook report, told Latitude Media that the projects originally planned two to three years ago predated the absolute acceleration in AI demand and today's labor and equipment shortages.

What Is Actually Causing the Slip

Three constraints are doing most of the work. The first is electrical equipment availability. Wood Mackenzie's most recent supply chain analysis shows lead times for large power transformers averaging 128 weeks and generator step-up units at 144 weeks. Pad-mounted three-phase distribution transformer shortages are projected to worsen in 2026 due to surging demand from data centers, manufacturing reshoring, and EV charging deployment. Power transformer prices are up roughly 77 percent since 2019. Manufacturers have announced approximately $1.8 billion in new North American capacity, but most of those plants will not reach commercial output until 2027 or 2028.

The second is grid interconnection. Power constraints can take up to five years to resolve in PJM, MISO, and parts of ERCOT. The interconnection queue does not move at the pace of AI capital deployment. The third is community opposition. Sightline cited the cancellation of a proposed $1 billion data center in Michigan, which a local official linked to Meta. Developers withdrew the rezoning application in December after months of public opposition over water use, grid strain, and lack of transparency. Similar moratoriums are now in place or proposed in at least twelve states. Community resistance has become what Wang described as a true material driver of attrition in the development pipeline.

The Hyperscaler Response Is Already Visible

Hyperscalers are not waiting for the constraints to clear. They are pivoting to behind-the-meter generation, on-site power, and hybrid architectures that allow campuses to operate independently of the regional transmission system. New Era Energy & Digital's 7 GW project in New Mexico is large enough to justify its own generation. Project Hummingbird in Greene County, Pennsylvania, will run on 944 MW of behind-the-meter natural gas turbines. Google acquired Intersect Power earlier this year, securing a direct energy lifeline rather than continuing to compete for grid capacity. While on-site and hybrid powering models account for less than 10 percent of total project count, they represent nearly half of announced capacity, driven by a small number of gigascale, grid-independent campuses.

Microsoft on April 30 disclosed via Data Center Knowledge a substantial gap between contracted AI workload demand and currently delivered capacity, driving accelerated procurement across third-party colocation, behind-the-meter partnerships, and direct power purchase agreements. The disclosure quantifies what the broader market has been observing: hyperscaler demand is no longer constrained by capital availability but by physical delivery of power, transformers, and water.

What This Means for the Building Owner

The first implication runs through industrial real estate near hyperscaler hubs. Sites in Northern Virginia, the Phoenix metro, central Ohio, and the Dallas-Fort Worth corridor were priced through 2024 and 2025 on assumptions that included credible grid access for new large loads. Those assumptions no longer hold for projects breaking ground in 2026. Industrial parcels that already have utility-confirmed capacity, ready interconnection rights, and transformer commitments now command a premium that did not exist eighteen months ago. Parcels without those features face longer entitlement and energization timelines that translate directly into delayed rent commencement and compressed yields.

The second implication runs through commercial real estate adjacent to delayed projects. The conventional thesis around data center campus development assumes the campus draws economic activity, supply-chain tenants, hospitality demand, and workforce housing. Delayed campuses do not deliver those secondary effects on the original timeline. Property owners who underwrote acquisition or development pricing on the basis of an announced 2026 campus opening date now need to extend their absorption schedules by 12 to 24 months in many cases.

The third implication is structural. Sightline Climate identified at least 9 outright cancellations against 777 tracked projects, suggesting that delays rather than full cancellations are the dominant pattern. But Bloomberg's reporting found that approximately half of US data centers planned for this year are expected to be delayed or canceled, with shortages of electrical equipment cited as a leading reason. The stretched pipeline functions as a slow-motion repricing event for any commercial real estate asset whose business case depended on a specific project energizing on a specific date.

The Speed-to-Power Premium

The structural insight from the Sightline data is that physical assets, energized power, transformers, interconnection rights, and built-out shells, are now the scarce factors of production for the AI buildout. Capital is abundant: the four largest hyperscalers have committed approximately $650 billion to AI infrastructure across 2025 and 2026. The constraint is on the delivery side. That inversion changes how commercial real estate should be underwritten.

Properties that can deliver speed-to-power, defined as utility-confirmed capacity, ready interconnection, available transformer capacity, and a credible behind-the-meter generation option, now command pricing that reflects the scarcity of the package. The June 30, 2026 sunset of Section 179D and the broader Inflation Reduction Act incentive stack reinforce the timing pressure. Projects that begin construction before the deadline capture a federal incentive package worth 40 to 50 percent of project cost. Projects that wait past the deadline carry a permanently higher net cost and a longer payback. The two pressures, energy delivery scarcity and the federal deadline, are operating in the same direction on the same calendar.

The Building Owner's Action Window

For commercial property owners not directly involved in the AI buildout, the implications are still concrete. First, any commercial property dependent on grid power for mission-critical load (cold storage, data center colocation, manufacturing, healthcare) is now operating in a market where new grid capacity is multi-year away and existing capacity has competing claimants. The case for behind-the-meter generation, battery storage, and controllable load has hardened from optional to defensive. Second, any commercial property with onsite generation, storage, or load-flexibility capacity has gained leverage in capacity market participation, demand response programs, and tenant lease pricing. Third, the legislative response is accelerating. Twelve states have filed statewide moratorium bills, North Carolina is considering 40 MW thresholds for cost-based tariffs, and Wisconsin's recent rewrite of the We Energies VLC tariff signals that the regulatory environment for large new loads is tightening, not loosening.

The Sightline 7 GW shortfall is not abstract. It is the operational expression of a grid that cannot keep up with the pace of demand growth, a manufacturing base that cannot keep up with the pace of orders, and a permitting process that cannot keep up with the pace of community pushback. For commercial real estate owners, the shortfall is a forcing function. Properties positioned for energy independence are repricing upward. Properties dependent on grid-delivered capacity at the original announced timelines are repricing downward. The repricing is happening now, on the 2026 cycle, not in some abstract future state.

That is the underwriting frame for May 2026. The grid is constrained. The hyperscalers are pivoting to behind-the-meter generation. The federal incentive window is 56 days from closing. Commercial real estate decisions made between now and June 30 will shape the next 25 years of asset performance.

Sources

Sightline Climate data center pipeline7 gigawatt shortfalltransformer lead timeshyperscaler behind-the-meter generationspeed-to-power premium
Back to Blog