Grids strain to keep pace as AI load queues explode and cost fights begin

Six months after US federal regulators ordered grid operators to show how they would speed up connections for data centers and other large loads, none of them have a full answer. A new analysis published by Utility Dive on October 8, 2026 found that power system planners are still missing near-term fixes while they negotiate the bigger, slower reforms. The gap between how fast demand is arriving and how fast grids can absorb it is now the defining problem in transmission planning, in the United States and in Europe.

Queues are growing faster than anyone can study them

Power system planners, including utilities, grid operators and regulators, are racing to develop new ways to bring large loads online faster, but they are often overlooking solutions that could increase capacity quickly while longer, structural remedies are negotiated. Shortages of skilled labor and grid equipment are slowing the buildout of new generation to meet rising demand from AI data centers, while market reforms must go through a regulatory process that can be prolonged. Many large loads are waiting to interconnect while Congress debates proposed legislation that would affect them, and in the meantime local governments and states are going their own way.

On June 18, 2026, the Federal Energy Regulatory Commission issued “show cause” orders to six system operators directing them to submit plans for speeding the interconnection of large loads. All six requested more time to develop processes for bringing these loads online without passing the costs on to other customers or threatening system reliability. Zachary Zimmerman, director of research and policy at Grid Strategies, told Utility Dive that “the elements of Order 1920 showed the way, and large loads add the exclamation point on the need for new planning processes,” but “there has not been a complete rethink of utility planning yet.”

Nowhere is the scale clearer than in Texas. ERCOT’s large-load interconnection requests went from roughly 63 gigawatts in late 2024 to 226 gigawatts a year later, driven mostly by data centers. By August 2026 the queue had reached about 474 gigawatts, with roughly 90 percent of the new power requests coming from data centers, according to a letter from Texas Governor Greg Abbott. That is more than five times Texas’ record peak electricity demand for ERCOT, Abbott wrote, and a BNEF analysis cited by Power magazine put ERCOT’s actual summer peak at about 91.3 gigawatts, meaning the interconnection queue is five times the size of peak demand.

ERCOT large-load interconnection queue growth

ERCOT had already tried to get ahead of the backlog. In June 2026 it rolled out “Batch Zero,” a framework to study large-load requests of 75 megawatts or more in groups rather than one at a time. ERCOT President and CEO Pablo Vegas called it “a fundamental shift in how ERCOT manages the significant growth of large load interconnection, providing a structured, transparent path forward.” That process stalled on August 3, 2026, when Abbott ordered a statewide audit of every data center project in the queue. In response, ERCOT postponed the Batch Zero transmission planning study, a framework approved in June 2026 to manage data center power demands through a batch evaluation process; applications had been due July 10, 2026. Industry analysis from BloombergNEF, reported by Power magazine and summarized elsewhere, estimated that the freeze put roughly a fifth of the entire US data-center development pipeline at risk of delay, with billions of dollars of exposure for developers.

Fast-track lanes open, but friction follows

In the PJM Interconnection, which spans 13 Mid-Atlantic and Midwest states, the response has taken a different shape: a temporary express lane for shovel-ready generation. On October 8, 2026, PJM approved grid interconnections for two energy storage projects planned by Engie IR Holdings, an 860-megawatt project in Crawford, Pennsylvania, and an 800-megawatt project in Morrow, Ohio, plus a 455-megawatt gas-fired uprate by LS Power at its Hunterstown plant in Adams, Pennsylvania, with all three expected online by mid-2029.

The Expedited Interconnection Track, approved by FERC in June 2026, allows PJM to consider up to 10 interconnection requests a year for new or uprated capacity resources larger than 250 megawatts that can come online within three years, sponsored by a state in PJM’s footprint. The process is set to expire at the end of 2027. It did not go unopposed: PJM’s plan faced significant challenges at FERC, with protests filed by clean energy trade groups, the Illinois Commerce Commission, LS Power, the New Jersey Bureau of Public Utilities, Vistra, and community organizations. FERC rejected those objections, saying the readiness criteria were “designed to attract ‘shovel-ready’ projects that have the highest probability of successfully proceeding through this limited interconnection process, regardless of whether they are proposed by an independent power producer or affiliated with an LSE.”

The urgency behind the fast track is visible in PJM’s capacity market. PJM’s wholesale power costs jumped 54 percent in one year, and the grid operator’s last two base capacity auctions came in short of its reserve margin targets, according to its market monitor, which said prices will keep rising until large data center loads are addressed. That is also why the Department of Energy is pushing PJM to tighten its cost rules before more large loads are approved. On October 7, 2026, the DOE urged PJM to quickly revise its reliability backstop procurement plan to prevent cost shifts onto existing customers from data centers and other large loads, and told PJM to file a proposal by October 29 in response to FERC’s finding that the plan’s cost-allocation framework needed changes.

Who pays: Europe reaches the same argument

The cost question is not confined to the United States. On October 7, 2026, RWE chief executive Markus Krebber told Bloomberg that “if you have more demand, you need more grid build out and you need more firm capacity,” and “this is currently not paid by those who are causing the problem.” He argued “we need to look at who bears the cost of the grid build out required to meet this additional demand, rather than spreading those costs across all consumers.”

The issue is becoming more pressing as Europe seeks to attract investment in artificial intelligence and data centers while simultaneously spending heavily to overhaul its electricity system. Krebber pointed to the US as a preview: data-center development is far more advanced there, and the boom has drawn opposition amid concerns that households and businesses are subsidizing the infrastructure required to power AI through higher electricity bills. He also flagged a market effect beyond grid fees: data centers are willing to pay more for power, which could put pressure on consumers and energy-intensive industries competing for the same supply.

Academic research is starting to quantify that spillover. A June 2026 working paper from MIT’s Center for Energy and Environmental Policy Research found that data center entry between 2010 and 2024 increased average US retail electricity prices by 2.7 percent, with effects of 2.1 percent for residential customers, 2.8 percent for commercial customers and 4.2 percent for industrial customers. These effects were concentrated among investor-owned utilities, where average retail prices rose by 5.6 percent, while they were much smaller among publicly-owned utilities and absent among cooperatives; price increases were also larger in states with deregulated electricity generation. NPR reported a concrete example: by one estimate, Maryland customers pay somewhere around $168 to $216 more per year on their electricity bills primarily due to the data center boom.

Some policy groups see an opportunity rather than only a burden. A Searchlight Institute report, covered by Grist, noted that US utilities make roughly $35 billion in investments in transmission infrastructure every year, far short of what’s actually needed, as electricity demand is projected to double or triple in the next 25 years. The report proposed a dedicated grid infrastructure fund to accelerate expansion, under which hyperscalers would pay into the fund in exchange for speedy connections. So far the response has been fragmented rather than coordinated: more than 30 US states have proposed or implemented measures of this sort.

California tries to force follow-through

Even where regulators approve transmission spending, construction often lags. A November 2025 report from the California Public Utilities Commission found that nearly 22 gigawatts of renewable generation and battery installations depend on transmission projects that have already experienced a delay in coming online, and of those, 13.2 gigawatts have already been delayed or are at risk of delay due to delayed transmission project timelines.

The state knows it needs more lines: California’s grid operator has approved just under $28 billion in new transmission projects over the past five years, but the major utilities are lagging on the follow-through. Meeting state climate targets will require adding 7 to 8 gigawatts of capacity a year for the next two decades, roughly double the current pace of growth.

California Governor Gavin Newsom has now signed legislation meant to close that gap by holding utilities and independent transmission owners accountable for delays, with regulators empowered to order “remedial actions” when projects fall behind. Alex Jackson, executive director of American Clean Power-California, which co-sponsored the bill, said the goal is not punitive: “the goal is not to punish utilities, but simply to provide more oversight,” he said. “Every part of the state’s energy planning system depends on utilities getting these projects done on time.”

The next bottleneck is hardware, not paperwork

Even where interconnection reform works, a second constraint is emerging behind it. A Nexans analysis published October 8, 2026 argues that a grid connection offer is a right to build, not a delivered substation, and not the same as getting power. Interconnection queues have stretched to years in some of Europe’s core hubs, and in the UK the queue nearly tripled in seven months before an emergency reform, yet even that reform only clears one link in the chain between an offer and energization.

The equipment behind that chain has its own lead times. According to the Nexans analysis, medium-voltage switchgear now averages around 44 weeks, with data center-class specifications running higher: standard switchboards near 52 weeks, power circuit breaker switchboards past 84 weeks, and medium-voltage gear approaching two to three years in some configurations. Large power transformers average well over two years, generator step-up units longer still, and lead times on high-voltage units from tier-one manufacturers have been reported as high as five years. The consequence is already visible in construction data: Bloomberg reported that of the near 12 gigawatts of US data center capacity slated to come online in 2026, only about a third was under active construction, with a shortage of transformers, switchgear and batteries cited as a cause.

Who gains, who loses

Developers that win a spot in a fast-track process, like Engie and LS Power in PJM, gain years of schedule certainty that ordinary queue positions cannot offer. Independent power producers and clean energy developers waiting in the regular queue argue the opposite: expedited lanes, by design, let a handful of projects jump ahead of others that have been waiting longer, which is exactly what LS Power and Vistra argued in their protests against PJM’s process even as LS Power simultaneously used the same track to advance its own gas uprate.

Ratepayers are the group with the least say and the most exposure. The MIT research and the Maryland estimate cited by NPR both point to the same mechanism: when cost allocation is not resolved before large loads are approved, a share of the bill tends to land on existing residential and commercial customers, especially in states with investor-owned, deregulated utilities. That is precisely the dynamic RWE’s Krebber is trying to head off in Europe before it becomes entrenched.

Jurisdiction	Queue or pipeline (GW)	Data center share	2026 action
ERCOT (Texas)	474	~90%	Batch Zero frozen for state audit, August 3
PJM (13 states)	n/a (fast-track: 2.1)	Majority of forecast load growth	Expedited Interconnection Track approvals, October 8
CAISO (California)	22 (delayed clean energy)	n/a	New accountability law signed by Gov. Newsom
Germany / RWE	n/a	n/a	CEO calls for cost reallocation to large loads, October 7

Risks and open questions

The biggest uncertainty is whether the demand behind these queues is real. ERCOT’s own leadership and the state’s utility regulator have both acknowledged that a queue five times larger than peak demand likely overstates how much will actually be built, since speculative requests from developers chasing multiple sites at once inflate the totals. If a large share of announced data centers get delayed, cancelled or relocated, some of the generation and transmission being rushed into service could end up underused, a cost that would also fall on customers.

A second risk is sequencing. Fast-tracking generation interconnection does not by itself solve the equipment and labor shortages the Nexans analysis describes, and clearing a queue without securing transformers, switchgear and skilled crews simply shifts the bottleneck downstream. A third is political: Texas shows that a state can halt a multi-year regulatory process overnight, which adds a layer of policy risk on top of the usual engineering and permitting timelines that developers and utilities already have to manage.

What to watch

PJM’s revised cost-allocation filing to FERC, due October 29, 2026, in response to the DOE’s intervention on large-load backstop procurement.
Whether Texas completes its ERCOT data center audit and reopens Batch Zero, and how many of the 474 gigawatts of requests survive verification.
Compliance filings from the six grid operators under FERC’s June 2026 show-cause orders, and whether any propose concrete near-term fixes rather than only long-term planning reforms.
How California’s new transmission accountability law is enforced against the state’s investor-owned utilities over the next reporting cycle.
Whether European regulators pick up RWE’s call to shift grid costs toward large loads, and how that interacts with ongoing grid-fee reviews in Germany and elsewhere.