AI Data Center Construction Falls for First Time Since 2020 as Power Constraints Bite

AI Data Center Construction Hits First Decline Since 2020

Construction of new AI data center capacity in the US fell to 5.99 gigawatts at the end of 2025 from 6.35 gigawatts at the end of 2024, marking the first decline since 2020 despite surging demand for artificial intelligence computing capacity1

1

. The downturn signals that the physical and political limits of the power system are starting to constrain the AI infrastructure buildout that has been powering US economic growth over the last 18 months3

3

.

Real estate brokerage CBRE Group Inc. reported that data center development faces mounting delays in permitting, zoning, and power procurement, forcing developers to look beyond traditional hubs like Northern Virginia, where construction underway fell 29%1

1

. Construction also dropped 15% in Hillsboro, Oregon, and 14% in Silicon Valley, while projects soared 169% in Chicago and 15% in Dallas-Fort Worth as developers seek markets with available land and power1

1

.

Local Opposition Creates Mounting Roadblocks for AI Boom

Community opposition to data center development has intensified dramatically, with 20 projects representing approximately $96 billion in investments blocked or delayed in the second quarter of 2025 alone2

2

. The pushback has accelerated further, with 10 new moratorium proposals filed in the past month across states including New York, Michigan, Virginia, and Oklahoma4

4

. Wilmer Hale reports that over 200 bills were introduced in 2025 aimed at regulating the sector across states, with 40 becoming law2

2

.

Grassroots protests against AI have focused on rising energy bills, noise nuisance, water usage, and concerns about AI eliminating jobs2

2

. Illinois Governor JB Pritzker sought to temporarily halt tax incentives for data centers to contain soaring power costs, while an Oracle Corp. site in New Mexico that secured tax incentives and government-backed bonds has prompted protests largely focused on environmental impact1

1

. MacroEdge research identified 26 project cancellations through January, up from just one in October3

3

.

Power Constraints and Grid Capacity Challenges

Power grid limitations represent the most significant barrier to AI data center expansion. Connecting to the energy grid is "the No 1 challenge we're seeing," said Marsden Hanna, head of energy and sustainability for Google, at a utility industry conference3

3

. Utilities in many markets are telling developers it will take four or five, sometimes 10 years to interconnect, with one utility informing Google it would take 12 years just to study the interconnection timeline3

3

.

The interconnection process can take regional grid operators up to five years to review how proposed power generation sources could affect the grid, creating a substantial bottleneck3

3

. Data center capacity is already under severe stress, with JLL noting that the sector only had 1% vacancies last year, while the overall vacancy rate in primary markets fell to a record low 1.4% at year-end2

2

1

1

.

Supply Chain Disruptions Compound Infrastructure Setbacks

Energy shortages combine with severe supply chain disruptions to create additional infrastructure setbacks. The surge in data center projects has strained energy grid supply chains still recovering from Covid-era snags and stressed by the clean energy buildout3

3

. High-voltage transformers that once had lead times of six months now take up to four years to manufacture3

3

. Only one plant in the US produces the type of steel needed for many grid equipment pieces, while copper shortages and skilled labor shortages represent additional constraints3

3

.

Up to 11 gigawatts of 2026 capacity "remains in the announced stage with no signs of construction," according to Sightline Climate4

4

. With typical build times of 12 to 18 months, that capacity could still come online but only with dramatic acceleration4

4

. More than one-quarter of the 110 data center projects slated to come online last year faced permitting delays4

4

.

Economic Implications and Future Outlook

AI demand is forecast to require more than $3 trillion in data center investment, including related power supplies, according to estimates from Morgan Stanley and Moody's Ratings1

1

. New tenants absorbed a record 2.5 gigawatts in 2025, up 38% from a year earlier1

1

. Nearly six gigawatts came online last year, and five gigawatts are already under construction this year, with one gigawatt capable of powering about 1 million US homes4

4

.

Proposed regulatory efforts are considering ways to address community concerns while allowing data centers to operate, potentially pushing builds further from cities, requiring developers to pay for their own grid and power updates, or mandating their own power sources2

2

. These zoning restrictions could increase compute cost to the industry as energy and compliance bills rise, while boosting demand for fiber, cooling pipes, and extra transmission lines2

2

. The complex knot of issues raises questions about the US's ability to quickly facilitate the data center boom, with potential broader economic implications as the datacenter machine that has been powering growth winds down3

3

.