WASHINGTON, 19 AUG 2026 — On 17 August the deadline expired on an order that will shape how data centres connect to the American grid. In June, the Federal Energy Regulatory Commission gave all six of the country's grid operators 60 days to justify their rules for connecting very large electricity loads, or replace them.
The proceedings cover territories serving roughly 200 million people across more than 30 states. The commission's decision will become the template for regulators elsewhere.
The orders
The commission acted under section 206 of the Federal Power Act, the provision that lets it examine whether existing tariffs are just and reasonable and require changes if they are not. Show-cause orders of this kind put the burden on the operator: demonstrate that your current rules work, or file replacements.
Large loads were defined for the purpose as those with a peak demand above 50 megawatts interconnecting at above 69 kilovolts, with loads above 20 megawatts falling within scope of the inquiry. Operators were also required to file informational reports on generation adequacy by 20 July, and several sought to have proceedings held in abeyance in early August.
The problem the orders describe
Grid interconnection rules were written for two kinds of customer: generators selling power, and load growing by a percentage or two a year. A single facility asking for several hundred megawatts on a two-year timeline fits neither.
The specific difficulty is that a large load arrives faster than the transmission and generation needed to serve it. Building a data centre takes eighteen months to three years. Building the substation, the lines and the generating capacity behind it takes five to fifteen. When the queue is full of requests on the shorter timeline, the operator has to decide who gets served, in what order, and who pays for the reinforcement — and most existing tariffs do not say.
The consequences of not answering are visible in the interim. Some operators have effectively paused new large-load connections; others have processed them first-come-first-served, which rewards whoever filed the paperwork earliest rather than whoever the system can actually accommodate.
Who pays is the question underneath
Every technical question resolves into one of allocation, which is what makes the issue contentious rather than purely administrative.
If a data centre triggers a transmission upgrade, the cost can be assigned to that customer, spread across all customers in the zone, or split. Assign it entirely and you have a large, lumpy charge that may push the facility to another state. Socialise it and ordinary households pay for infrastructure built to serve one industrial customer. Neither is obviously correct, and the choice is a political one dressed as a rate design.
The second allocation question is about reliability, a topic operators are often reluctant to discuss. Under scarcity, someone is curtailed first. Large loads with flexible operations — training runs that can pause, workloads that can shift — are natural candidates, and tariffs that pay for that flexibility rather than merely permitting it are among the more promising replacements likely to be filed.
We examined this directly in our reporting on data-centre tariff design, and the American proceedings are the largest test yet of the same question.
Why this matters in Southeast Asia
While these are American rules for American grids, they represent the first serious attempt by a major regulator to write large-load rules from scratch. Regulators in Southeast Asia face the same arithmetic on a shorter runway.
Malaysia, Indonesia, Thailand and the Philippines have all seen data centre announcements whose combined demand is significant relative to national generating capacity, and none of them has the reserve margin that would let the question be deferred. Johor's cluster alone has pushed peninsular Malaysia's planners into revising demand forecasts, and Thailand has already begun conditioning investment approvals on grid capacity — we reported on the Board of Investment requiring an ERC letter before approval.
The difference is procedural rather than substantive. American operators are being made to publish their rules and defend them in a public docket, with intervenors, comment periods and a reasoned decision at the end. Regional decisions of the same kind are more often made through investment approvals, ministerial discretion and bilateral negotiation with individual developers, where the trade-offs are settled but not written down.
That matters for anyone building here, because an unwritten rule is one that can change. A developer who negotiated a connection on favourable terms in 2024 has no assurance that the same terms apply in 2027, and the transparent-docket approach produces something a fifteen-year capital commitment can be underwritten against.
What the responses will reveal
The filings made against the 17 August deadline will differ from one another in ways that reveal each operator's priorities.
Operators in constrained territories with the largest data centre pipelines have the strongest incentive to propose restrictive terms — high contributions in aid of construction, firm curtailment obligations, or minimum-take commitments that protect other customers if the facility does not materialise. Operators with spare capacity have less reason to change anything and may argue their existing tariffs are adequate. The show-cause procedure is designed to test exactly that kind of claim.
The abeyance requests filed in early August are their own signal. A request to pause suggests an operator negotiating with stakeholders rather than defending the status quo, which usually precedes a settlement rather than a contested order.
What we could not establish
What was actually filed. The 17 August deadline has only just passed and the individual responses had not been reviewed at the time of writing, so it is not possible to say which operators defended their tariffs, which proposed replacements, or how far the proposals differ.
Also unestablished: which abeyance requests were granted; the commission's expected timeline for decisions; whether it intends to impose a uniform framework or accept regionally distinct answers; how the interim treatment of queued requests works while proceedings are open; and what the 20 July generation-adequacy reports showed.
What to watch
The first question is whether the commission pushes for a common framework or lets six different answers stand. A uniform rule would be the template other regulators could copy.
Then watch the cost-allocation language specifically. Whatever formula emerges for splitting network reinforcement between the large load and the general body of ratepayers will be quoted in tariff proceedings far outside the United States, because every regulator facing this question needs a defensible precedent and none of them wants to invent one.
Finally, watch where the projects go. If the resulting terms are materially stricter in one territory than another, data centres will be sited accordingly. That is how a domestic American story becomes relevant to planners in Johor, Batam and Laguna.
View author profile → · Editorial policy
