Analysis
The Grid Bottleneck: Why Data Centers Can't Get Connected
Explosive demand from AI-driven data centers has collided with a power grid that was never designed to absorb hundreds of gigawatts of new, always-on load. The result is a backlog of interconnection requests so large it now dwarfs the grid itself — and a multi-year wait that is reshaping how the built environment gets financed, sited, and built.
None of this is a rumor or a single utility's growing pains. It's a pattern documented consistently by Lawrence Berkeley National Laboratory, the Federal Energy Regulatory Commission, ERCOT, and PJM Interconnection — the institutions that actually run and regulate the grid. Together, their data tells a clear story: demand for grid access is growing far faster than the grid's physical and administrative capacity to deliver it, and data centers are now the single biggest driver of that gap.
01 · The Backlog
A queue larger than the grid it wants to join
Nationally, Lawrence Berkeley National Laboratory's "Queued Up" tracker counted roughly 2,061 GW of generation and storage capacity across some 8,200 projects actively waiting for transmission interconnection at the end of 2025 — a figure that is roughly one-and-a-half to two times the entire installed generating capacity of the United States. That total actually fell about 10% from its 2023 peak of ~2,600 GW, but not because the system got faster. It fell mainly because a wave of speculative projects withdrew.
That national queue tracks power generation. The newer and, in many ways, more alarming backlog is on the demand side — the queue of large loads, overwhelmingly data centers, waiting to draw power from the grid. Nowhere is that clearer than in Texas.
In roughly a year, ERCOT's large-load queue grew from about 63 GW to more than 400 GW — a figure that now exceeds ERCOT's all-time peak electricity demand for the entire state, which sits around 85–86 GW. Put simply: Texas has more data-center load asking to connect than the grid has ever had to serve, period. Around 89–90% of that queue is data centers. Much of that growth traces back to how quickly rack-level power density has climbed as GPU clusters replace conventional servers — a trend we chart in detail in our GPU rack power density evolution chart (2020–2027).
The queue is not a reliable predictor of what actually gets built. Historically, only about 13% of generation capacity that requested interconnection between 2000 and 2020 had reached commercial operation by the end of 2025; roughly 75% was withdrawn, and about 10% remains active in the queue. Large-load queues are widely believed to carry a similar, if not larger, share of speculative or duplicate requests — developers reserving grid capacity at multiple sites before committing to one.
- Reached commercial operation ~13%
- Withdrawn from the queue ~75%
- Still active, undecided ~10%
02 · The Timeline Problem
Why "grid-ready" now means a multi-year wait
Connection timelines haven't just stayed slow — they've been getting steadily worse. For generation projects, the median span from interconnection request to commercial operation has stretched from under two years for projects built between 2000 and 2007, to more than four years for the 2018–2024 cohort, to more than five years for projects that actually reached completion in 2025.
PJM, the largest U.S. grid operator, shows the same pattern from a different angle: projects that entered service in 2025 took an average of more than seven years from start to finish. Roughly three of those years go to securing an interconnection service agreement; the remaining four go to what comes after approval — building substations, upgrading transmission, and waiting on equipment.
- Permitting 29%
- Supply chain (transformers, equipment) 23%
- Other cited factors 28%
- Remaining, not separately broken out ~20%
The physical bottlenecks are real and hard to shortcut. New high-voltage transmission lines routinely take five to ten years or more once permitting and construction are both counted. High-voltage transformers and other long-lead equipment — the literal hardware needed to energize a substation — now carry multi-year order backlogs, up from roughly two to two-and-a-half years just a few years ago.
The grid's approval process was built for generators feeding power in — not for hundred-megawatt, near-24/7 loads pulling power out. Data centers are asking an analog-era queue to behave like a real-time reservation system.
03 · Market Deep-Dive
The three markets defining the crisis
Grid congestion isn't uniform. It concentrates wherever hyperscale demand and legacy infrastructure meet head-on. Three markets illustrate the range of the problem — and the range of responses now underway. Because congestion, climate, incentives, and permitting speed all vary so much by state, matching a project to the right market has become as important as the technical design; our data center site selection criteria in the United States breaks down the factors worth weighing before you commit to a market.
ERCOT
438–474 GW
~89–90%
~85–86 GW
PJM
7+ years
~30 of 32 GW
~6.5–6.8 GW
Dominion / N. Virginia
~7 years
4–10 yrs
up to 12 yrs
ERCOT's answer: Batch Zero
Facing a queue five times larger than its own peak demand, the Public Utility Commission of Texas approved ERCOT's "Batch Zero" process in June 2026. Instead of studying each project one at a time, ERCOT now groups large-load requests of 75 MW or more for joint study, applies financial-security requirements to weed out non-serious applicants, and sorts projects into base, studied, and excluded classifications.
PUCT approves Batch Zero
Joint-study framework for large loads ≥75 MW, with financial-security requirements designed to filter speculative projects.
First notifications expected
Projects learn their classification — base, studied, or excluded — determining their path forward.
Transmission plan due
ERCOT's resulting transmission build-out plan is expected, translating the batch study into concrete projects.
PJM, meanwhile, is moving to give large loads "conditional" status above 50 MW, consider prioritizing uncommitted large loads for curtailment during tight periods, and add registry requirements — changes aimed at keeping the queue honest and protecting reliability for existing customers.
04 · Root Causes
Five reasons the grid can't keep up
No spare headroom
The grid lacks the transmission capacity and firm generation to absorb near-24/7, multi-hundred-megawatt or gigawatt-scale loads on top of existing demand.
The wrong process for the job
Serial, first-come-first-served studies were designed for generators injecting power, not massive loads withdrawing it — and speculative "phantom" projects keep triggering costly restudies.
Physical build times
New high-voltage transmission can take five to ten-plus years once permitting and construction are combined; transformers and other long-lead gear now carry multi-year backlogs.
Supply chain and permitting
In PJM, permitting and supply-chain issues together account for over half of post-approval delay — the phase after a project is already technically "approved."
Who pays for upgrades
Regulators are wrestling with cost allocation — how much of the transmission and generation buildout gets billed to the new large load versus spread across existing ratepayers.
05 · The Federal Response
FERC steps in
On June 18, 2026, the Federal Energy Regulatory Commission issued tailored Section 206 "show-cause" orders to all six jurisdictional grid operators — PJM, MISO, SPP, CAISO, ISO-NE, and NYISO. Each must justify its existing interconnection tariff or propose reforms within about 60 days, covering faster study processes, transparent cost-shifting rules, treatment of co-located generation, and flexible service options for large loads. Resource-adequacy reports are required as well, all aimed at speeding up large-load integration — data centers and manufacturing alike — without compromising reliability or shifting costs unfairly onto other customers.
These orders build on FERC's earlier Order 2023 generator-queue reforms and run alongside regional experiments already underway: co-location of data centers directly at generation sites, behind-the-meter generation that bypasses the public grid entirely, and non-firm or flexible-load service agreements that let a data center connect sooner in exchange for accepting curtailment during system stress.
Advisory
What developers, investors, and site selectors should do now
Grid interconnection risk has become underwriting risk. Treating it as a late-stage engineering detail, rather than an early-stage investment variable, is the single most common — and most costly — mistake in today's data center and industrial real estate market. A practical playbook:
- Verify queue position before you underwrite the site, not after. Request the interconnection study status and queue number directly from the utility or RTO; a signed letter of intent is not the same as a signed interconnection agreement. Run each candidate parcel against a structured data center site selection checklist before it goes into your pipeline.
- Price a 5–10 year connection timeline into your base case, not a best case. Model the downside scenario where transmission upgrades slip and financing carry costs accordingly. Our data center power calculator is a useful starting point for sizing the load you'll actually need to reserve in the queue.
- Evaluate behind-the-meter and co-located generation early. On-site gas, nuclear, or renewables paired with storage can bypass years of queue time — but bring their own permitting and financing complexity that needs its own diligence track.
- Diversify across RTOs and utilities. ERCOT, PJM, and other markets carry different congestion profiles, regulatory postures, and reform timelines; a multi-market pipeline reduces single-point regulatory exposure.
- Track the FERC dockets and state proceedings directly. Section 206 responses, ERCOT's Batch Zero classifications, and PJM's large-load reforms will materially change which sites move to the front of the line over the next 12–24 months.
- Consider flexible-load and curtailable contracts where the business case allows it — accepting occasional curtailment in exchange for a materially faster interconnection date can beat waiting in the standard queue. Modeling facility efficiency with a PUE calculator helps translate a curtailment scenario into real operating impact before you sign.
- Build utility relationships before you need them. Markets moving to batch studies and financial-security requirements reward applicants who engage early, provide real project maturity signals, and avoid the "phantom project" label that triggers scrutiny.
Core Insights Review contributors publish research-based analysis and editorial insights on commercial real estate, PropTech, smart infrastructure, sustainable construction, industrial real estate, and emerging technologies shaping the future of the built environment.
Check for more information: Core Insights Review
