No Reservations ⚡

The power grid is done holding capacity on a promise. FERC and Texas moved the same day to bring the rules in line with a fast-evolving power markets

By Bryan J. Kaus

For a successful technology, reality must take precedence over public relations, for nature cannot be fooled. - Richard Feynman

On Thursday, June 18, the Federal Energy Regulatory Commission ordered the six grid operators under its jurisdiction to defend or rewrite the rules that govern how data centers and other large customers connect to the power system. The same day, in Austin, the Public Utility Commission of Texas approved a new process for its own grid to study large loads as a group rather than one at a time.

Texas runs its grid the way it does in order to stay outside federal jurisdiction. That is the entire point of ERCOT. So when Washington and Austin reach for the same lever on the same day, with no coordination and no common boss, it tells you something worth hearing. The thing they are both responding to is not a question of who holds authority. It is a question of physics, and of arithmetic.

The arithmetic is the part worth sitting with.

What FERC actually did

The Commission’s tool was a show-cause order under Section 206 of the Federal Power Act, issued to PJM, MISO, the Southwest Power Pool, the California ISO, ISO New England, and the New York ISO. Each operator has sixty days to justify why its current tariffs remain just and reasonable, or to propose changes. Each has thirty days to file a separate report explaining how it will keep enough generation on the system to serve the load it is being asked to connect. Together these regions serve roughly two hundred million people.

Strip away the procedural language and the order is about one thing. Cost. The Commission flagged several areas for reform, and the load-bearing one is the prevention of cost-shifting. It directed each region to show how it keeps the expense of connecting a hyperscale data center on the party that requested it, rather than on the household three counties over. It pointed toward arrangements that would have developers post money up front, so that a project which never gets built does not leave its bill behind for everyone else to pay.

This is the right place to slow down. The fight over speed-to-power is usually framed as an engineering problem. How fast can we energize. The more honest framing is an allocation problem. The cost of underbuilding is paid by the company that waited. The cost of overbuilding is paid by everyone else. A developer who reserves five hundred megawatts and builds fifty has not made a private mistake. He has consumed studies, planning, and upgrade commitments that the system priced in and that someone now has to absorb. Speed-to-power is not free. The question is never only how fast. It is how fast, for whom, and on whose bill.

The Texas number

Texas gives you that arithmetic in its rawest form. ERCOT defines a large load as a single site that draws more than seventy-five megawatts, enough to power something like nineteen thousand homes on a peak afternoon. It is currently tracking more than 438 gigawatts of such requests. Nearly nine in ten of those gigawatts are data centers.

Hold that figure against the grid it wants to join. ERCOT’s all-time peak demand, the most electricity Texas has ever pulled at once on the hottest August evening in its history, is about 85.5 gigawatts. The request queue is five times the largest load the system has ever served.

It is not real. Not all of it. A request to join a queue is close to free, and a free option gets exercised early and often. The same project gets registered in three places. A site is reserved against a financing round that has not closed. When Enverus screened the Texas queue project by project, it estimated that when this first round settles, something on the order of twenty-two gigawatts will actually qualify as ready. Against 438 requested. The rest is intention, not load.

That gap is what the new process exists to find. Rather than study each request in the order it arrived, ERCOT will study them together and sort them into three buckets: projects mature enough to count, projects that still have to prove themselves, and projects that do not make the cut. The old rule was first come, first served. The new rule is closer to first prove you are real. Reserving capacity will require site control, financial security, and a credible commitment to energize, not a place in line and a hope.

A free queue measures appetite. Batch Zero measures what is real.

The same frontier, approached from two sides

Look closely and the federal order and the Texas process are wrestling with the identical move. A large load that cannot get through the queue fast enough has an obvious workaround. Bring your own generation and sit next to it, drawing less from the grid or nothing at all. Co-location, in the language of the filings.

This is the part the old tariffs never contemplated. A data center parked beside its own gas plant or behind its own meter is neither a normal load nor a normal generator, and the rules were written for one or the other. FERC’s order spends much of its length on exactly this, directing operators to write clear terms for co-located generation and behind-the-meter arrangements. In Texas, Enverus counts roughly a dozen gigawatts of projects that have already skipped the queue this way. Both grids arrived at the same frontier from opposite directions. The load is trying to route around the constraint, and the regulator is trying to write the rule before the routing becomes the norm.

Not a step behind

It is easy to read all of this as regulators caught flat-footed, and easier still to make it a morality tale about government always lagging the market. That reading is lazy. The interconnection rules now being rewritten were sound for the world they were built for, a world where load grew a percent or two a year and the hard problem was connecting generators, not customers. The data center did not break the rules. It changed the magnitude of the thing the rules were measuring, and it did so faster than any planning cycle is designed to absorb.

What you are watching is a system retrofitting itself in public. FERC opened its first show-cause on co-located load in early 2025, ordered PJM to fix its tariff at the end of that year, and widened the order to all six operators six months later. ERCOT spent the same period discovering that a process built to study requests one at a time could not survive a single year in which it received more large-load requests than in the three years before it combined. Neither moved early. Both moved. Adapting to a change of this size is not an indictment of the institution. It is the work.

The Point Taken

The takeaway is simpler than the headlines. A request to connect is not a reservation. The grid spent years treating the two as the same and quietly putting the cost of the gap on everyone’s bill. On June 18, FERC and Texas began, separately and on the same day, to move that cost back onto the party making the request. What follows is a fight about cost allocation, not technology, and it gets decided in the fine print.

Underneath it sits one question. When a new kind of customer arrives wanting power at a scale and speed the system was never built to deliver, who carries the risk that the customer is wrong about itself?

The old answer, by default and without anyone choosing it, was the ratepayer. The connection got built, the load came in light or never came at all, and the cost stayed on the system. What FERC and Texas did on the same Thursday was begin, clumsily and incompletely, to move that risk back toward the party making the claim. Prove the load is real. Post the security. Carry the cost of your own optimism.

That is not hostility to growth. It is the precondition for it. A grid that cannot tell a real megawatt from a hopeful one will either choke on caution or drown in cost, and neither one builds anything. The decisive work of the next two years lives in the details, the penalty rates and the security postings and the cost-recovery terms that the financial engineers are modeling right now. That is where it gets settled whether speed-to-power is a privilege the system extends to those who can prove they will use it, or a subsidy it quietly hands to those who cannot.

The queue is full of real intention, much of it serious. But you do not build steel and copper against a hope, and a grid that tries will always overshoot or fall short. Make the promise pay to prove itself, and you build enough for the demand that is real.

© 2026 23.5 Strategies and The Point Taken™