Show-cause orders are different from rulemakings. A rulemaking invites comment and takes months. A show-cause order says: justify this, or it changes. FERC used Section 206 of the Federal Power Act because Section 206 gives the agency authority to act when it finds existing rates may be unjust or unreasonable. The agency doesn’t have to wait for a complaint. It can initiate on its own motion.
That’s what happened on June 18. According to AP reporting, FERC simultaneously ordered all six major US regional transmission organizations, PJM, MISO, SPP, CAISO, ISO New England, and NYISO, to address the same fundamental question: are your current tariffs for large-load customers just and reasonable given the scale and speed of AI data center interconnection demand? The docket is RM26-4-000.
The action reportedly followed a request by Energy Secretary Chris Wright in 2025. Wright’s role is confirmed; the specific request hasn’t been independently verified from primary sources in as of publication, but it’s consistent with the DOE’s stated posture on data center power supply.
What Section 206 Actually Requires
The show-cause mechanism has teeth. Operators must file responses within 60 days, by August 17, 2026. Those responses must either: (a) demonstrate that existing tariff structures appropriately allocate interconnection costs for large-load customers, or (b) propose specific reforms that would make them appropriate.
If an operator fails to file a satisfactory response, FERC can impose reforms unilaterally. That’s the “or else” that makes this more than a policy signal.
Five reform categories are reportedly under consideration, according to reporting on the orders, including co-location arrangements, behind-the-meter generation, flexible transmission products, large-load cost allocation, and ratepayer protections. A 20 MW capacity threshold has been cited in reporting as the likely trigger for large-load classification, though this specific figure hasn’t been confirmed from primary sources.
The Central Dispute: Who Pays for the Wires
The cost-allocation fight is where this gets concrete for data center operators.
When a 500-megawatt hyperscaler campus connects to the transmission grid, it requires significant network upgrades: substation expansion, new transmission lines, protection equipment. Under current tariff structures in some regions, a portion of those upgrade costs gets socialized across the existing customer base through transmission rates. The hyperscaler gets fast interconnection; residential and commercial customers pay incrementally higher utility bills to support the infrastructure that makes it possible.
FERC’s stated aim, per reporting, is to prevent that cost-shifting. The commission wants large-load customers, specifically AI data centers, to bear the costs they impose on the grid rather than distributing those costs to retail ratepayers who have no stake in AI infrastructure.
Warning
The 20 MW threshold cited in reporting as the large-load trigger has not been confirmed from primary sources. Operators' actual proposed thresholds may differ. Data center operators with planned facilities in the 15–30 MW range should not assume their projects fall below the reform scope until operator filings specify the threshold.
Who This Affects
For data center operators and hyperscalers, this means: the economics of grid-connected campuses are about to become clearer, and potentially more expensive. The era of socializing interconnection costs may be ending at the federal level.
Regional Variation: Where the Reform Pressure Is Highest
The six operators face different situations.
PJM serves 65 million people across 13 states and DC, the largest grid operator in the world by load. It also has the largest backlog of queued interconnection projects in the US, a significant portion of which are AI data center connections in Virginia, Ohio, and Pennsylvania. PJM’s response to Docket RM26-4-000 will effectively define what “reform proposal” means in practice, because every other operator will calibrate against it.
CAISO operates California’s grid under a different set of constraints. California has its own renewable portfolio standards, its own Public Utilities Commission, and its own history of state-level disputes with FERC over jurisdictional boundaries. CAISO’s response may include a jurisdictional challenge, an argument that FERC’s Section 206 action encroaches on state ratemaking authority in ways that California is entitled to contest. That’s a slower path, and one that could produce a legal challenge rather than a reform filing.
MISO (Midwest), SPP (Central US), ISO New England, and NYISO each have regional characteristics, MISO’s coal transition dynamics, NYISO’s offshore wind interconnection backlog, but none carries the structural complexity or jurisdictional complication that PJM and CAISO introduce.
The Ratepayer Protection Fight
NARUC, the association of state utility regulators, has historically pushed back on federal authority when it perceives encroachment on state ratemaking. This action is a candidate for that pushback. State regulators can argue that retail rate design, including how transmission costs are allocated across customer classes, is a state-level function that Section 206 doesn’t fully displace.
That argument has lost in federal courts before. But it’s a plausible legal theory, and a coordinated NARUC intervention filed during the 60-day response window could extend the timeline or introduce a parallel proceeding in a federal circuit court.
Three Scenarios by August 17
What to Watch
*Scenario 1: Operators justify current tariffs.* Low probability. FERC’s order language implies the commission has already formed a preliminary view that existing structures are problematic. An operator that files a pure justification without reform proposals is essentially inviting FERC to impose changes unilaterally.
*Scenario 2: Operators propose targeted reforms.* Most likely path. Expect PJM to file a detailed reform proposal that establishes a cost allocation framework for loads above a defined threshold (likely near the reported 20 MW figure). Other operators will likely follow similar structures. This produces a new tariff regime within 6-12 months of the August filing, subject to FERC approval.
*Scenario 3: Legal challenge delays the process.* Plausible, particularly from California. A state regulator or utility filing for emergency stay in the DC Circuit could pause the response deadline while courts consider the jurisdictional question. This is a delay tactic, not a resolution, but it could push the practical outcome into 2027.
What Enterprise Teams Should Do Now
The 60-day window is not passive. Congressional bills on AI data center power have been moving on a separate track, the FERC action doesn’t wait for legislation, and it doesn’t require it. The practical steps for enterprise AI infrastructure teams:
First, identify which RTO or ISO covers your current and planned data center locations. The reform proposal from that operator will be the governing document for interconnection economics in that region. Second, follow the RM26-4-000 docket filings directly, FERC’s docket system is public, and the operator filings will contain more operational specificity than any press coverage. Third, if you’re evaluating new data center locations, the behind-the-meter generation option (co-locating a generator rather than connecting at the grid boundary) may become more attractive depending on how large-load cost allocation reforms are structured, model that scenario now rather than after the tariff structure is set.
TJS synthesis: The August 17 deadline is when the policy uncertainty about AI data center grid costs begins to resolve into specific rules. Teams that track the RM26-4-000 filings will know before their competitors whether the interconnection economics for their planned sites are about to change materially. FERC doesn’t move fast by default, the gigawatt race for US power capacity has been accelerating regardless of regulatory pace, but show-cause orders under Section 206 are one of the few mechanisms that create a hard deadline. This one is real.