The largest buildout in a generation is stalling at the connection point. That’s where the supplier opportunity is
Jamie R. Van Doren
This is the second in a three-part Trend Watch series on what’s actually changing in energy. Part one covered the largest power customers in the country walking away from the utility model. This installment covers what stopped them anyway.
On August 3, the governor of Texas told the data center industry to stop.
Less than a year after branding his state the epicenter of AI development, Greg Abbott ordered a halt to data center approvals. No new connections to the Texas grid until the Public Utility Commission of Texas and ERCOT audit what’s already sitting in the queue. Developers now have to disclose the tax breaks they’re receiving, how much power they’ll draw and generate, how much water they’ll use and how they’ll cool the facility, what they’re doing to limit the effect on the people who live nearby, and who actually owns the building.
His instruction to regulators wasn’t subtle. Any project that fails the process has to be denied.
The number behind the order explains it. ERCOT’s interconnection queue holds roughly 474 gigawatts of requests. That’s more than five times the state’s all-time record peak demand. Around 90 percent of it is data centers.
Texas wasn’t even first. New York paused approvals for large data centers in July, for up to a year, while it writes new rules. ERCOT has already postponed the first round of classification notices under its new large-load process and goes back to state regulators on August 20 for more time.
They Solved Supply. It Didn’t Help.
Last month we covered the largest power consumers in the country deciding they could no longer wait for the grid. Microsoft restarting Three Mile Island. Meta buying the entire output of an Illinois nuclear plant. They looked at utility timelines, decided the math didn’t work, and went and bought the megawatts themselves.
It worked, on paper. Meta’s Illinois output starts in 2027. Three Mile Island targets 2028.
Those are contracts, not electrons. The buildings meant to use that power still have to be permitted, connected, cooled, and backed up. Some of them can’t be.
That’s the constraint hiding inside the energy conversation, and it isn’t generation. It’s delivery. Getting electricity from where it’s made to where it’s needed, through equipment that exists, into a building that can shed the heat, backed by something that turns on when the grid can’t.
Each of those is a separate point of failure. Each of them is somebody’s contract.
The Wall Is Made of Equipment and Permission
Start with the hardware. GE Vernova closed the second quarter with 116 gigawatts of gas turbine orders and slot reservations, up from 100 three months earlier, and is now booking 2031 delivery slots. Siemens and Mitsubishi are four to five years out too. That isn’t a shortage of demand, capital, or willingness to build. There’s a shortage of the physical objects that move and convert electricity, and of the factories that make them.
Then permission. Interconnection studies, transmission planning, environmental review, local zoning. A transmission line in this country still takes ten to fifteen years to permit and build. The Texas order didn’t create that friction. It formalized it.
Then heat. A gigawatt-scale computing facility is, in engineering terms, a very large space heater with a business model attached. It’s a determining constraint on where the project can go at all, which is why water use showed up in a grid reliability order.
And then behavior. Grid planners historically treated large industrial loads as steady and predictable. AI compute isn’t.
On July 10, 2024, a lightning arrestor failed on a 230 kV line in northern Virginia. Auto-reclose settings turned that into six faults in 82 seconds, and roughly 1,500 megawatts of data center load came off the system at once. Those facilities are built to protect themselves from voltage disturbances. NERC formed a task force.
Three weeks ago it happened again at double the scale. On July 22, a line fault in Ashburn pushed more than three gigawatts onto backup power, about three percent of PJM’s demand. Dominion was clear it hadn’t shed those customers. Their own protection systems did. The disturbance registered as far west as Chicago.
Nothing was short of power. The load moved faster than the system could respond, because equipment inside private buildings decided it should.
The Grid Is Already Borrowing From Private Buildings

In May, the Department of Energy issued an emergency order under Section 202(c) of the Federal Power Act authorizing PJM to deploy backup generation resources at data centers and other major facilities across the Mid-Atlantic. In late July, a comparable order covering the Southwest Power Pool allowed operators to call up idle backup generation across 17 states, in territory home to about 20 million people. In early July, PJM issued the first emergency backup-generator warning in its history.
These orders usually get read as a story about how close the grid is running to the edge. That’s true. It’s also not the most useful part.
The useful part is where that reserve capacity actually lives. It lives in private buildings. It’s the generator set behind a data center, the switchgear in a hospital basement, the controls package that decides what happens in the four seconds after utility power drops. Federal reliability policy now quietly assumes all of that equipment exists, works, has fuel, and answers when called.
Somebody specified that equipment. Somebody installed it, integrated the controls, negotiated the fuel contract, ran the load bank test, and shows up when it fails. Those are supplier relationships, and the reliability of the American grid is now leaning on them.
Ohio Is Already Paying For This
None of this is happening somewhere else. Ohio sits inside PJM, the grid operator covering thirteen states, and PJM’s capacity market is the clearest scoreboard we have for whether delivery is keeping up.
It isn’t. PJM has now missed its own reliability requirement in two consecutive capacity auctions, the first time in its history the entire region has come up short. July’s auction fell 6,831 megawatts short and cleared at the federally approved cap of $325 per megawatt-day, for $16.4 billion. PJM says it will ask federal regulators in September for a special backstop procurement. In the prior auction, nearly 5,100 megawatts of the increase in forecast peak load came from data centers.
Closer to home, the numbers are contested. Regulators approved an increase to AEP Ohio’s transmission rider in March that added about $7.90 a month for a typical household, though offsetting decreases elsewhere on the bill left the average net change closer to twenty cents. Under AEP Ohio’s large-load tariff, data centers must pay for at least 85 percent of the capacity they reserve whether they use it or not. The Ohio Manufacturers’ Association has appealed that tariff to the Ohio Supreme Court, and a bill in the legislature would extend it statewide.
If you operate a facility in Ohio, this is already in your budget. If you sell to companies that operate facilities in Ohio, this is already in theirs.
Delivery Is a Procurement Category
This is the part most people miss.
You don’t have to generate electricity to sell into the energy buildout. The work is in everything that stands between a megawatt and a running operation.
Electrical contracting at industrial scale. Substation and switchgear work. Mechanical, HVAC, and thermal management. Water treatment and cooling systems. Civil and site construction. Controls, instrumentation, and automation. Backup generation, load bank testing, and commissioning. Fuel supply and emergency logistics. Operational technology cybersecurity, a different discipline from IT security. Environmental and permitting consulting. Ongoing operations and maintenance.
Some of that work has already started. In Virginia, owners are retuning facility control systems so sites ride through brief faults instead of dropping offline, and federal regulators have told NERC to write mandatory standards for large computing loads by year end. All of it has to be engineered, installed, tested, and documented, site by site.
For an MBE, the question isn’t “Are we an energy company?” Almost none of these firms are. The question is “Do we work on the systems that make power usable?” For a surprising number of businesses in our network, the honest answer is yes, and has been for years, in buildings nobody called an energy project.

The New Rules Are Also a New Opening
Look again at what Texas is asking developers to submit. Tax breaks. Power use and generation. Water use and cooling. Efforts to reduce the impact on nearby residents. Facility ownership.
Three of those are engineering questions. Two aren’t.
Community impact and ownership disclosure are social license questions, and they just landed inside a grid connection application. That happened because the politics moved. On July 18, a group called HumansFirst coordinated 142 protests across 42 states against the data center buildout. A June Reuters/Ipsos poll found only about a third of Americans approve of the current pace of construction, and just 14 percent would want a facility in their own community.
For a buyer assembling a project team under that kind of scrutiny, a supplier base that’s local, certified, documented, and visibly connected to the community stops being a values statement and starts being part of the filing.
No regulator has said anything about supplier inclusion. But when reviewers start asking who benefits from a project and who owns it, the answers get easier for buyers who can point to real local firms doing real scopes of work. That’s a case OMSDC’s members are unusually well positioned to help make, and almost nobody is framing it that way yet.
What To Do About It
For MBEs:
Rewrite your capability statement in the language of this market. Not “general contracting” but substation civil work. Not “IT services” but OT security assessment for industrial control systems. Buyers in this sector search for specific scopes, and generic language reads as no experience. Then get clear about which scopes you can carry alone and which require teaming. The pursuit pod model from our Growth Under Pressure white paper applies directly. One firm rarely covers a delivery scope. Three often can.
For corporate members:
Your supplier bench is about to be tested by demand you probably didn’t forecast. Waiting for a formal RFP means competing for what’s left. Start the qualification conversations before the constraint becomes an emergency, and look hard at the maintenance, backup, and commissioning categories. Those are the ones that quietly determine whether your operation stays up the next time a fault ripples through the system.
The Takeaway
For a century, the hard question in energy was how to make more of it. That question got solved. What they found on the other side was a second wall, built out of transformers, permits, cooling towers, interconnection queues, and public patience.
The next decade of energy work isn’t going to be dominated by the companies that produce power. It’s going to be dominated by the ones that deliver it, cool it, back it up, secure it, and keep it running when the margin gets thin.
That’s not an energy industry. That’s a construction, trades, engineering, logistics, and services industry, and it’s the one our members are already in.
Next month we close the series with what gets built next, and with a company right here in Ohio that’s already building it.
Until then, the question to sit with is this. If the biggest buyers in the country can own the power and still not be able to use it, who do you think they’re going to call?


