.svg){kind=avatar}
A few years ago, data center load growth was a line item in a planning study—important, but orderly. A curve to track, a number for the planning department to refine in due course. It was a problem you scheduled.
That era has closed. Large load is no longer something the grid quietly absorbs on the way to a forecast; it has become the thing the grid has to ration. And the rules now being written to do that rationing are doing something more consequential than they first appear. On the surface, regulators are cleaning speculators out of the interconnection queue. Underneath, they are quietly reassigning the job of building power—from the utility to the customer. The large load is no longer just asking the grid for electricity. It is increasingly being told to show up as its own power developer.
That shift is the real story, and in this issue of the Green Molecules® Journal we follow it through three markets—ERCOT, PJM, and FERC—to where it ultimately points: a wave of demand that the grid can no longer fully serve, redirected toward distributed generation, firm power, and the molecule infrastructure built to carry it.
For two decades, interconnection reform meant one thing: generators waiting in line. Solar, wind, and storage piled into study queues faster than grid operators could evaluate them, and the entire policy apparatus was built to triage that supply-side backlog. Load, by contrast, was assumed. You served it. It arrived gradually, it was geographically diffuse, and it rarely needed a queue of its own.
Texas has ended that assumption. ERCOT is now tracking more than 438 GW of large-load interconnection requests, and roughly 89% of that volume, by megawatt, is data centers. The scale is difficult to overstate: ERCOT's all-time system peak sits near 86 GW, which means the pipeline of requested new load is several times the size of the grid operating today. The trajectory is near-vertical—at the end of 2024 the large-load queue stood near 63 GW, meaning it has multiplied roughly sevenfold in about eighteen months.
No one believes all 438 GW is real. That is precisely the problem. A grid operator cannot build for demand it cannot trust, and it cannot ignore demand that might be trusted. A single gigawatt-scale facility can reshape the study results for every other project on its corridor, so the speculative requests are not harmless placeholders—they distort the planning picture for the real ones. When the queue can no longer tell a funded hyperscale campus from a reserved option, a process built for a smaller, slower era starts to strain against a problem it was never designed for. Everything that follows is regulators moving quickly to rebuild that signal.
Texas's response is the clearest signal of where this is heading, and it is worth reading closely, because other markets will borrow from it.
In June 2026, the Public Utility Commission of Texas approved ERCOT's "Batch Zero" framework—a deliberate break from one-at-a-time review toward coordinated batch studies of large loads of 75 MW or greater. The new standards, codified as 16 TAC §25.194, are built to force a load to prove it is real before it can hold a place in line:
This is usually described as a filter, and it is. But look at what the filter is actually demanding. A $50 million deposit per gigawatt, hard site control, and—written directly into the Texas framework—an explicit pathway for large loads to develop their own onsite generation and to let ERCOT curtail them when the grid is tight. Read together, these are not the terms you set for a customer. They are the terms you set for a co-developer. The regulation is screening out speculators by raising the bar to one that only an entity prepared to act like a power project can clear. First-ready is not just separating real from fake. It is redefining what a "real" load looks like—one that arrives with capital, commitment, and increasingly its own supply.
If Texas shows how the bar is being raised, PJM shows why regulators feel they have no choice: the cost of getting this wrong is already on the bill.
According to PJM's independent market monitor, data center load accounted for roughly $16.6 billion in capacity-auction revenue across the grid operator's last two auctions—about half of the total. That is not a forecast of someday; it is cost already locked in, flowing to customers across thirteen states and the District of Columbia. And it is accelerating: in the first quarter of 2026, PJM's all-in wholesale power cost rose 75.5% year over year, from roughly $77.78 to $136.53 per megawatt-hour, with the monitor attributing current capacity-market conditions almost entirely to large data center additions, both real and forecast.
Here is the mechanism that makes this a fight. The capacity market was built for organic load growth—households, businesses, the gradual electrification of the economy. What it is absorbing instead is concentrated blocks of demand arriving faster than firm supply can be built. When that demand lifts the clearing price, it lifts it for everyone: a household pays more because a hyperscaler's forecast tightened the auction. That dynamic is politically unsustainable, and regulators know it. The monitor's own prescription is telling—it has recommended that large loads be required to bring their own generation, with faster interconnection for those that do.
Notice that this is the same answer Texas reached, arrived at from the opposite direction. Texas is raising the entry bar; PJM is trying to stop socializing the cost. Both roads end at the same place: the large load is expected to supply more of its own power. When the two largest organized markets in the country independently conclude that the customer should bring generation, that is no longer a queue-management tweak. It is the new shape of the market.
The clearest sign of how fast this is moving came on June 18, 2026, when FERC issued six "show cause" orders directing the country's federally regulated grid operators to justify or rewrite their rules for connecting large loads. The move put data center interconnection squarely under federal review. But the deeper federal effort is older and more structural. FERC Order No. 1920 is the attempt to get ahead of demand rather than react to it, and setting the two policies side by side exposes a tension worth naming. Energy Capital Ventures® has engaged this question directly when former FERC Chairman Willie Phillips joined a recent Limited Partner Roundtable for a fireside chat on the impact of data centers and the evolving regulatory relationship.
Order 1920 requires transmission providers to conduct long-term regional planning at least once every five years, across a minimum 20-year window, using at least three forward-looking scenarios. Projects are tested against a defined set of benefits, must clear a benefit-cost screen, and arrive with cost-allocation methods agreed in advance. The rule also presses providers to "right-size" replacement facilities—when a line is being rebuilt anyway, build it for the demand you can already see coming.
The instinct is sound: build the backbone before the crisis, not after. But the federal and state approaches are working on different clocks. Federal policy looks twenty years out and plans transmission ahead of load. State policy, closer to the cost, asks load to prove itself before new build is committed. Both are right on their own terms—yet they are not yet reconciled, because it is hard to size two decades of transmission around demand that is still being screened for reality. 438 GW of requests against an 86 GW peak is a forecast you have to take seriously and cannot fully build against at the same time. So the forward-looking plan and the prove-it-first screen meet in a gap measured in years: years to permit, years to study, years to energize. That gap is time the largest loads do not have.
A planning system deliberately built to be careful, serving customers who need power on an AI buildout's timeline, leaves a space in the middle. That space is where power gets built outside the grid's critical path.
Put the three markets together and a single conclusion emerges. The demand is real, the grid cannot fully serve it on time, and the rules now actively push the largest loads to bring their own power. That redirected demand does not vanish. It flows toward generation and infrastructure that can be sited with the load and energized on the load's schedule.
This is the natural gas supply chain that Energy Capital Ventures® was built to back. When a gigawatt-scale load is required to post capital, hold its site, and increasingly supply itself, the practical answer is dispatchable, energy-dense, infrastructure-compatible power—onsite and firm gas generation, fuel cells, the methane and carbon management that comes with new build, and the efficiency and optimization layers that stretch every molecule further. The regulation is not slowing this demand; it is channeling it into distributed infrastructure, and into the Green Molecules® technologies that ride existing pipelines and sites rather than waiting on new grid corridors.
We have said before that the market increasingly rewards what can be deployed, not just what can be imagined. The new rules of power access put a sharper point on it: the loads that get built are the ones that show up with their own answer to power. For an early-stage fund positioned across that supply chain, the policy shift is not a headwind to navigate. It is the demand signal.
