The Federal Energy Regulatory Commission has issued show cause orders to all six major U.S. grid operators, giving them 60 days to either defend or revise their tariffs for connecting large loads above 50 MW, according to a commentary published June 2025 in Utility Dive. The June 18 action under Section 206 of the Federal Power Act targets PJM, MISO, SPP, CAISO, NYISO and ISO-NE, finding their current tariffs inadequate because they lack clear and consistent rules for large loads like data centers. FERC defined a large load as peak demand above 50 MW connecting above 69 kV, and ordered each operator to report within 30 days on how it will ensure enough generation exists to serve these loads.
The orders demand reforms across five areas: a real interconnection process for large loads, cost transparency so new loads don't shift network-upgrade costs onto existing customers, clear rules for co-location and behind-the-meter generation, a new class of transmission service for loads that can flex, and a study process for generation sited next to those loads. The action traces back to an October 2025 letter from the Energy Secretary pushing FERC to standardize large-load interconnection. Interconnection queues were built for a world where the big new thing connecting to the grid was a power plant, not a 300-MW load, and the process shows it—a single large load at 100 to 500 MW now lands on a system that was never designed to evaluate it quickly.
According to Shalin Savalia, a senior electrical engineer at Amazon Web Services writing in his personal capacity, "the most expensive thing on a data center campus is the one nobody photographs: a finished building that can't turn on." A campus can sit like that for months, and in the worst regions for years, with hundreds of millions of dollars in capital waiting on an interconnection study that grinds forward on a timeline no one in the building controls. The same company can develop three campuses in three regions and face three completely different processes, three cost-allocation methods and three timelines, with no national standard for how a large load connects. Savalia writes that in several states, data centers are tied to the majority of projected load growth, and the political backlash has followed with rate cases, moratorium proposals and local opposition.
On the inconsistency problem, forcing all six operators to answer the same questions at the same time is the first real attempt to pull large-load rules toward a common standard, which would let a developer expect the same basic playbook in PJM as in MISO and remove a real source of delay and risk. The cost-transparency push is easy to read as a threat but smarter to read as a shield—if a tariff makes it clear that a data center is funding its own network upgrades and not leaning on the neighbors, the local rate case argument gets much weaker. On flexibility and co-location, FERC is explicitly asking operators to create service for loads that can flex and to write real rules for bringing your own generation, treating co-location as a solution rather than a threat to be studied into oblivion.
But this is a show cause order, not a final rule—the opening of a proceeding, not the end of one. The grid operators get 60 days to respond, then come their filings, protests, technical conferences and likely rehearing requests, and nothing about June 18 energizes a single stranded building next quarter. Savalia argues the fastest path to power now runs through things data centers already control: every hyperscale campus is ringed with on-site generation that idles most of the year, and their cooling loops, redundancy and ability to shift workloads give them far more room to curtail than they admit in an interconnection meeting. The developers who show up with an offer—how many megawatts they can drop, for how long, how quickly, and how much generation they bring themselves—will turn FERC's framework into actual energized racks, while the ones who wait for the old model to get faster will still be staring at a finished building that can't turn on.