Two weeks ago, the Trump administration and a bipartisan group of mid-Atlantic governors urged PJM to call for an emergency procurement auction, arguing that new baseload supply should be built quickly to accommodate data center power demand.
The move came as a surprise to most, but it reflected an increasingly urgent reality. Data center growth is outpacing grid capacity. BloombergNEF projects that U.S. data center power demand will reach 106 GW by 2035, more than double current levels. But meanwhile, there is potential capacity to spare. The challenge isn’t a shortage of proposed projects; it’s getting them built. Across the country, planned generation is in regulatory limbo or stuck waiting for long-lead equipment.
On orders from the president, several major offshore wind projects sit frozen off the East Coast, waiting for courts to decide their fate. Offshore wind developers are facing mounting financial pressure, and several gigawatts of contracted capacity are in limbo.
At first blush, these two storylines don’t seem related. One’s a demand problem; the other’s supply. But they’re both stuck on the same thing: build throughput, or how fast the grid can connect and deliver new projects — and one of the metrics that will define the U.S. market in 2026.
Demand side: Data centers
The AI era is turning data centers from background noise into a system-planning challenge.
Large artificial intelligence data centers hit the grid as a demand shock. New campuses often need network upgrades, new substations, and transmission-level interconnection, rather than using existing capacity. And nowhere is this more visible than in PJM. The grid operator’s January 2026 outlook projects that summer peak demand will rise to approximately 222 GW by 2036, roughly 66 GW above current levels.
Some data center projects use behind-the-meter generation as a workaround; they pair data centers with existing generation to bypass constrained interconnections, as seen with Amazon’s campus connected to Talen’s Susquehanna nuclear plant. However, these setups still need firm capacity, fuel, and at least some connection to an already-stressed grid.
Political opposition typically emerges during the siting process. Water use, local tax benefits, and “who pays” disputes between hyperscalers and ratepayers have all become significant political flashpoints that sometimes make national news. This drama is partly why the Department of Energy asked FERC to develop rules for large load interconnection. The resulting rulemaking process (known as RM26-4) is ongoing, with dozens of commenters weighing in on standards for loads of 20 MW or more.
Supply side: Offshore wind
Offshore wind is a mirror image: a clean energy supply shock that runs into many of the same bottlenecks as data centers, just on the generation side. Projects depend on federal permits and their resilience to litigation, specialized ports and vessels, and onshore interconnections into congested coastal grids.
Like data centers, they also encounter local opposition, and changes in federal policy can halt progress regardless of project economics. In the U.S., several projects awaiting permit approvals found that their original offtake agreements, signed years prior, were no longer economical. This led to cancelled PPAs and supply chain uncertainty.
The Department of the Interior’s late-2025 stop-work order on five major East Coast projects — Vineyard Wind, Revolution Wind, Coastal Virginia Offshore Wind, Sunrise Wind, and Empire Wind — demonstrated how a single federal decision can halt gigawatts of contracted capacity and waste billions in investment across multiple states.
While early 2026 court decisions have allowed Revolution Wind and Empire Wind to resume installation, these cases highlight how heavily installation timelines now depend on the judicial system. Offshore wind’s long-term potential is still large, but its near-term delivery schedule will depend on the pace of permitting offices and courts.
The shared stress: Build throughput
Data centers and offshore wind hit the same wall: The grid can only connect so much, so fast. Large transformer lead times are between 80 and 120 weeks, with some transmission-class units taking three to six years. Even when financing is lined up, equipment availability is the bottleneck that stalls projects.
Cost allocation adds another layer. FERC Order 2023 pushed generators to cluster studies, stricter readiness standards, and firm timelines, but disagreements over who should pay for shared network upgrades continue to stall decision-making. The outcome of RM26-4 will likely reopen the same question for big customers: How much of a hyperscaler’s upgrade bill should be tied to the site, and how much should be shared across the system?
Financing exists. Projects are ready. But permitting offices, interconnection queues, equipment manufacturers, and cost-allocation debates all move more slowly than the grid can afford to wait.
What happens when forecasts miss
In a baseline assumption, data center growth stays close to current forecasts, offshore wind projects advance, and the grid handles them both without major disruption. Queues stay long, congestion deepens in a few regions, and planners rely on solar, storage, and new gas to keep reserve margins intact. But what if that doesn’t happen?
Two scenarios break the baseline. In a “Load surprise”, data center demand rises faster than expected, and the grid can’t keep up. In a “Supply Shortfall”, offshore wind and other major generation projects get delayed, leaving the system short on the capacity it was counting on.
In either case, the system falls back on whatever can be built or connected to the grid the fastest.
Another scenario exists where permitting accelerates, equipment manufacturing scales up, and cost allocation is resolved. But 2026 is too soon for structural reform at that scale. Order 2023 and RM26-4 are steps in the right direction, but they won’t fix transformer lead times or eliminate litigation risk this year.
Resource deployment speed depends on the type of project. The scorecard below shows how different resources, large loads, and transmission face permitting delays, interconnection backlogs, equipment shortages, and local opposition. The more red flags a project collects, the longer it takes to deliver.
So what’s going to happen this year?
Given the constraints, it’s likely that developers will start to care less about megawatts on paper and care more about what can actually get built. In either a Load Surprise or Supply Shortfall, speed-to-power becomes the only metric that matters. Procurement will lean toward firm, dispatchable resources, with greater use of location-specific contacts and hedges against local congestion.
Offshore wind’s near-term path, meanwhile, may depend as much on court decisions as on weather conditions.
And policy is just trying to catch up. Order 2023 tries to fix the generator queue and RM26-4 on the load side. Both aim to make interconnection more predictable. But policy reform takes years, while data center siting takes months. This mismatch is turning the race from who can find the most power, to who can find the power that will actually turn on.
