Texas Data Centers ⚡ Power Permits on Hold: What’s Next
Texas data centers face power permits on hold as energy demand outpaces planning. Learn what’s next for ERCOT interconnection and behind-the-meter power.
Texas Data Centers: Why Power Permits Are on Hold
Texas Paused Data Center Permits. The Grid Problem Is Bigger Than AI
Texas has become one of the main places in the US to build data centers. Developers can find land, power infrastructure, and a business climate that attracts big tech companies. But the build-out is bumping into a hard constraint: the grid can't connect every proposed campus on the timeline companies would like.
The interconnection queue is the clearest signal. ERCOT, which runs most of Texas's electricity grid, reported 63 gigawatts (GW) of large new load requests at the end of 2024. By the next June, that number had climbed to 474 GW. About 90% of the requests were tied to data centers. Put simply, 474 GW is more than five times Texas's record peak demand.
Texas's response has included stricter requirements for new projects and a pause on permits while officials review the pipeline. People may frame this as an argument over AI, but it's really about something less headline-friendly: how to plan, fund, and run a grid reliably when demand is growing unusually fast. This article breaks down what makes a grid connection hard, what power a campus can generate itself (and where that doesn't solve the problem), and which changes could help Texas add capacity without pushing unreasonable costs onto households and businesses.
Why Texas hit pause
A long interconnection queue doesn't mean all those projects are real, ready-to-build commitments. Developers might file requests for multiple sites, secure a spot in line before they have a customer, or run early tests on whether a location could work. That can be sensible at the company-exploration stage - but at scale, it makes planning a headache.
Grid operators have to figure out what substations and transmission lines need upgrades, what those upgrades will cost, and how to decide which project pays for what. If a proposed campus never gets financing or a customer, the infrastructure planned for it may be unnecessary. And if costs get spread too widely, the people who are already connected can end up covering upgrades meant for projects that never arrive.
ERCOT was built to handle fewer large-load applications at once. As requests surged, Texas regulators approved a batch study process for large loads, including projects of 75 MW or more. Instead of treating each application in isolation, ERCOT can study groups of projects and assign the limited available grid capacity among them. It's a more scalable approach - but it still relies on knowing which proposals are serious.
The queue needs evidence, not just ambition
Texas has added deposits and other requirements meant to discourage speculative filings. That helps, but deposits can't prove a developer has a committed customer, secured equipment, financing, permits, or a construction schedule that's actually workable. Even a meaningful deposit is often small compared with what it costs to build a gigawatt-scale campus and the supporting infrastructure.
A credible process needs milestones tied to real progress. For example, a developer might pay for the capacity it reserves, fund its share of network upgrades, and hit phased construction targets. If it misses those milestones, the reserved capacity can be reassigned. That gives serious projects a way to plan while reducing the incentive to hold a place in line indefinitely.
The permit pause also reflects another concern: whether projects are being transparent with the communities that will host them. Residents want clear information about water use, noise, emissions, construction impacts, and how (or whether) local power bills might change. In the original analysis, described through a state survey, only 28 of 377 companies had responded by June. That doesn't exactly build trust.
Private site searches can be legitimate. Naming a major buyer too early can move land prices and complicate negotiations. But once local officials are asked to approve tax breaks or infrastructure commitments, it's harder to justify withholding basic details from the public.
So the pause is about more than whether Texas wants data centers. It's about setting rules that separate projects that can deliver from those that may not - and making sure local communities understand the trade-offs before commitments are locked in.
What connecting to the grid really involves
Buying electricity is not the same thing as reserving a physical path for it to reach a building. Electricity moves through a network, and what that network can handle depends on the grid's electrical conditions. A contract with a generator doesn't automatically reserve transmission capacity between that generator and a data center.
For years, a large customer could often work with a utility that already had room on its network, sign up as a large load, and contract for energy. That was easier when utilities had spare capacity. Today, a major new campus - or a big power plant - may need new substations, transformers, and transmission lines before it can operate at full scale.
Interconnection studies are reliability tests
An interconnection study isn't just paperwork. It looks at what happens if parts of the grid fail. Planners might model a transformer outage and then test whether the system can handle the loss of a line or generator, too. These checks matter because a data center is a large, concentrated load. The grid has to keep operating safely when equipment fails - not just on paper, but in real fault scenarios.
There's also a risk that comes from how data centers behave. Many facilities shift to backup power when they detect a voltage disturbance. If a large amount of load disconnects from the grid at nearly the same time, the resulting change can destabilize the system. Texas has adopted voltage ride-through requirements designed to keep new data centers connected through routine faults.
That matters even more for ERCOT, which operates a relatively self-contained grid with limited connections to other US regions. It can't rely on neighboring grids to absorb disturbances. Reliability rules can look overly strict when you're focused on a single project, but they're built around system-wide outcomes - including extreme weather, when demand is high and outages are expensive.
Who pays for the upgrades?
The challenge isn't only deciding which upgrades are needed. It's deciding who covers them. A new transmission line might serve a data center, but it can also improve reliability for existing customers or enable other future development. Those benefits are real, but it's hard to assign the costs precisely before the infrastructure is built.
Texas has often linked some transmission costs to customers' demand during summer peaks. That can encourage a large customer to reduce demand at critical times. But it also means a campus that needs a lot of capacity doesn't necessarily pay in a clean, direct way that matches the size of the connection it wants. Regulators have discussed methods that more closely reflect capacity reserved for a specific customer.
There's no perfect formula. If a new project pays too little, households can end up subsidizing infrastructure that benefits a private business. If it pays for every upgrade that might benefit other parts of the network, it may stall projects that could drive genuinely useful grid expansion. The goal is transparent cost allocation - clear responsibility for upgrades tied to a project, and a defensible way to share costs when upgrades provide wider benefits.
For more on ERCOT and its planning process, start with ERCOT's grid information. The basic idea is straightforward: an interconnection date is a promise to both a project and a shared system. That promise should be made only when the underlying assumptions hold up.
Why developers are building power on-site
When a grid connection could take five to ten years, waiting comes with real costs. Data center operators have expensive computing equipment that doesn't earn revenue until it can run. For many operators, speed to power isn't a nice-to-have - it's a business requirement.
One response is to build generation on-site. People often call it behind-the-meter power, but the goal is the same: start operating sooner, even if the full grid connection arrives later.
The common plan isn't necessarily to stay off-grid forever. A developer might use gas turbines, engines, fuel cells, batteries, or a mix of technologies to begin operations and then connect to the grid when capacity opens up. The on-site system may remain in place as backup or as a way to reduce how much the campus draws from the grid during tight periods.
That can work as a bridge, but bridges still need an end point. On-site power can be more expensive to run than grid electricity, and it doesn't remove the need for permits, fuel supply, equipment delivery, and a connection plan. Some campuses described as "off-grid" have also planned infrastructure for a future grid connection. That isn't automatically a contradiction - it can reflect a phased strategy - but the plan has to be stated clearly.
An islanded campus has to recreate grid services
A campus that runs independently can't rely on the rest of the grid to pick up the slack when a generator trips. It has to manage voltage, frequency, reserves, and the ability to recover from outages. Those services are handled by the grid in the connected case, and they aren't automatically "free" when a facility operates as an island.
AI workloads add another wrinkle. Training and inference can change electricity demand quickly. At one early xAI site in Memphis, swings of 10 to 20 MW several times per second reportedly stressed turbine equipment. Batteries can smooth those swings, but they add cost and require careful integration.
Fuel security matters too. Gas generation can provide power on demand, but reliability depends on whether the fuel pipeline is available. Winter Storm Uri showed how cold weather can disrupt multiple parts of both gas and electricity systems at once. Solar avoids fuel-delivery risk, but it's intermittent; carrying a gigawatt-scale campus through a long winter night would require a very large battery system.
There's also a difference between equipment reliability and campus reliability. Even if a power plant runs most of the time, it still has planned maintenance and unexpected outages. To deliver near-continuous power in islanded operation, a developer may need redundancy that looks closer to a second full power system. That means more equipment, more operational complexity, and more maintenance.
Some operators may accept a lower uptime target in exchange for getting online years earlier. That can be a rational trade if the workload and customer contracts can tolerate it. Still, it's not a magic shortcut - the business is trading reliability for speed, and the decision should be explicit.
Private power does not remove regulation
Generating electricity for your own campus puts you in a different regulatory position than selling power to other customers. In Texas, selling power to a neighbor or building a private network across a utility's service territory can trigger additional rules and approvals. A private grid serving multiple campuses may offer operational benefits, but legal frameworks still limit who can own and operate it.
Air permits have to match how generators will actually run. Equipment approved only for emergencies can't simply become everyday power supply just because a grid connection is delayed. Developers need to line up the operating plan, emissions limits, fuel arrangements, and interconnection strategy before construction starts.
The bottom line: on-site power can reduce dependence on the interconnection queue, but it doesn't make the project independent of infrastructure, regulation, or reliability engineering. It shifts the bottleneck rather than removing it.
How flexibility and new infrastructure could help
A data center doesn't have to be either a fixed, full-load customer or a fully isolated island. Some sites can reduce consumption during a grid emergency, switch to on-site generation, use batteries, or shift computing workloads to another time or location. That kind of flexibility can make a large project easier to accommodate.
But flexibility only helps if it's dependable and available when the grid needs it. A campus that promises to cut load but can't do so without breaking customer commitments isn't a reliable resource. Operators need controls, backup arrangements, and clear agreements about how quickly they can respond.
Texas's batch process includes options that let some projects bring generation or accept limits on how much grid power they can draw when the network is constrained. That can help some campuses connect sooner. Still, there's an important distinction: a flexible connection may not remove the need for planners to build upgrades for the full load the campus ultimately wants. Flexibility can improve operations without necessarily shrinking the infrastructure requirement.
Better use of existing equipment
Some upgrades can increase capacity or improve operators' visibility without waiting for an entirely new transmission corridor. Reconductoring replaces older wires with higher-capacity conductors. Dynamic line ratings estimate how much a line can carry based on current conditions like weather, rather than relying only on conservative fixed ratings. Advanced transformers and monitoring tools can help operators see and control power flows more precisely.
Batteries and other distributed resources may also relieve specific constraints. But grid planners need evidence that those resources are located correctly, can perform reliably, and can be dispatched when needed. If planners can't count on them during the planning process, they may still approve conventional infrastructure as a backup.
That's why practical pilots matter. The point of a pilot is to test flexible resources under real operating conditions, measure performance, and define rules for when planners can credit those resources toward capacity. A successful pilot is more useful than a theoretical claim because it gives operators something they can plan around.
The long-term buildout is a supply-chain problem
Even with better planning and more flexibility, the US still needs additional generation and transmission if demand continues to rise. A 100 GW load running year-round would consume about 876 terawatt-hours - roughly one-fifth of recent US annual electricity consumption. No single source or engineering trick can cover that scale quickly.
Gas is available in many markets and can provide firm power, but it comes with fuel and emissions risks. Solar and batteries can scale, but supply chains and weather-dependent output still matter. Some existing nuclear plants can sometimes increase output, and new reactors or geothermal could provide firm power over a longer horizon. Each option has different timelines, costs, and operational constraints.
The bottleneck isn't only generation. Large transformers can take years to deliver. Switchgear, inverters, skilled crews, permits, and transmission projects all have their own lead times. A project can have a site and a customer and still get stuck because a component or a line can't be built fast enough.
Texas can make the process more predictable by pairing milestones with enforceable commitments. Developers that fund their first phase, pay for reserved capacity, and meet construction targets shouldn't automatically be treated like speculative applicants. At the same time, utilities and builders should be held responsible when delays come from their side. Clear timelines and transparent cost rules are more useful than a blanket promise that everything will connect quickly.
The stakes go beyond AI. Factories, homes, electric vehicles, and other new loads will need much of the same equipment. Data centers are hitting constraints early because they're large and can move quickly, but they can also end up financing upgrades that other users need later. Whether that investment helps existing customers depends on who pays, how costs are allocated, and whether projects actually deliver what they promised.
Conclusion: make the fast lane accountable
Texas doesn't have to choose between endless data center growth and a blanket halt. What it needs is a process that filters out weak applications, protects reliability, makes developers pay for the capacity they reserve, and gives credible projects a timeline they can actually finance.
On-site generation and flexible demand can buy time, but they don't replace the need for a stronger grid and more power supply. The durable solution is a mix: credible interconnection rules, well-designed private power where it's workable, dependable flexibility, and real investment in generation, wires, and equipment.
If you're evaluating a data center, an energy project, or a grid technology, start with the constraint that really controls the schedule: connection capacity, equipment availability, permitting, fuel, or financing. Then check whether the project has a realistic plan for each one. That's a better starting point than assuming the grid - or a generator - will be ready when it's needed.