Data Center Energy and FERC's AI Load Order

Data Center Energy and FERC's AI Load Order Data Center Energy moved from a planning issue to a tariff and interconnection issue on June 18, 2026, when the Federal Energy Regulatory Commission used Section 206 of the Federal Power Act to issue show-cause orders to six regional grid operators. The action covered PJM, MISO, SPP, CAISO, ISO-NE, and NYISO, and it asked each operator to justify existing rules or propose reforms for data centers and other large loads. By September 7, 2026, the 30-day and 60-day windows described in the order had already passed. The research record supplied here does not include the filings or any final FERC decisions that followed those deadlines, so the careful reading is retrospective but limited: the order changed the regulatory process and forced data-center load questions into formal tariff review, but it did not itself establish a single national technical standard for AI facilities. Data Center Energy Rules Now Face FERC Review Why Section 206 Matters FERC's June 18 action required each covered grid operator to show why its existing tariff was just and reasonable, or to propose changes. The same action required informational reports within 30 days on available generation capacity for existing and new large loads, and gave the operators 60 days to respond on the tariff question. FERC described five areas of concern: transmission service applications and study processes, transparency of transmission costs, co-location with generation, behind-the-meter generation, and transmission service for flexible large loads, according to the agency's June 18 order announcement. The technical point is that FERC did not order a data center to use a specific cooling system, chip architecture, battery size, or workload scheduler. The order targeted grid-facing rules. That distinction matters because many AI facilities are designed around internal engineering constraints, including compute density, power distribution, backup generation, thermal management, and uptime targets. FERC's order instead focused on how those facilities request grid service, how costs are assigned, and whether flexible load structures can be treated differently from fixed industrial demand. What Data Center Energy Demand Means For Operators Grid operators have had to study large loads that can appear faster than traditional transmission planning cycles. AI training clusters and inference campuses can require high-capacity service, and the requested load may depend on construction timing, equipment delivery, customer contracts, and local generation options. FERC's order put these requests into a common regulatory frame, but the six operators still run different markets and planning processes. That variation is not a small administrative detail. PJM, MISO, SPP, CAISO, ISO-NE, and NYISO differ in resource mix, congestion patterns, queue design, capacity market structure, and state-policy interaction. A tariff change that fits one region may not fit another without shifting risk to generators, data-center customers, or other ratepayers. Data Center Energy planning, therefore, becomes a question of evidence: how much load is firm, how much can be curtailed, how much generation is dedicated, and how much transmission cost should follow the requesting customer. What Changed Technically In Large Load Interconnection Interconnection Is Not Only A Queue Problem The order treated large-load interconnection as more than a waiting line. The five reform areas point to specific operating questions. If a data center applies for transmission service, the grid operator needs to know whether the load will draw continuously, whether it can reduce demand during system stress, whether it is paired with a generator, and whether behind-the-meter generation reduces or increases grid risk. Each answer affects studies, upgrades, and reliability assumptions. Alternative transmission technologies were included in the study-process discussion. The research record does not identify which technologies each operator proposed or assessed, so it would be unsupported to claim that any single technology will solve the bottleneck. The more defensible interpretation is narrower: FERC wanted grid operators to examine whether existing study processes were too slow or too opaque for large loads whose connection requests can affect regional capacity needs. For Data Center Energy managers, the order raises documentation demands. A facility seeking interconnection may need clearer evidence about load ramping, backup systems, planned energization dates, and any capability to reduce demand. Flexible large-load service could become valuable if tariffs recognize it, but flexibility must be measurable and enforceable. A data center cannot claim flexibility only as a commercial label; the grid operator needs operational terms that can be dispatched, audited, and priced. Co-Location And Behind-The-Meter Limits Co-location with generation is one of the most contested areas because it can reduce strain on transmission in some cases while creating cost-allocation disputes in others. If a data center sits near or behind a generator, the facility may argue that it should pay less for network service. Other market participants may argue that shared grid services and reliability support still carry costs. FERC's order did not settle that question in the research record; it compelled the operators to explain or reform their treatment of those arrangements. Behind-the-meter generation also needs careful classification. On-site generation may help meet local load, yet it does not automatically eliminate transmission or reserve requirements. Fuel supply, outage coordination, emissions permits, synchronization, and protection settings can all affect reliability. In AI campuses with tight uptime targets, backup generation may be sized for continuity rather than regional support, which means grid operators still need evidence before treating it as dependable capacity. Cost Allocation And Ratepayer Exposure Why The PJM Auction Figure Matters The cost issue is not theoretical. A letter from Senator Edward Markey cited an estimate that PJM's 2024 capacity market auction cost $16.6 billion, compared with $2.2 billion in 2023, with increasing demand from large loads such as data centers identified as a driver and costs borne by roughly 65 million PJM ratepayers, according to the data center energy costs letter. That figure does not prove that AI data centers were the only cause of the auction increase. Capacity prices can be affected by supply retirements, reliability rules, demand forecasts, market design, and transmission limits. The figure is still relevant because it illustrates the stakes of tariff design. If large customers trigger new capacity or transmission needs, regulators must decide how much cost follows the customer and how much is socialized across the region. Policy Area Technical Question Cost Risk Transmission Studies Can studies capture large, fast-moving load requests? Delayed upgrades or overbuilt assets Cost Transparency Are upgrade costs visible before commitment? Unexpected charges or cross-subsidy claims Co-Location Does local generation reduce grid dependence? Disputes over network service fees Flexible Load Can load reductions be verified during stress? Reliability risk if flexibility is overstated Data Center Energy cost allocation therefore sits between engineering and finance. The equipment load is physical, but the tariff decides who pays for the infrastructure needed to serve it. Investors and operators gain more predictable project economics when upgrade costs are clearer. Ratepayers gain protection when tariffs require the customers driving incremental costs to pay an appropriate share. Operational Risks For AI Infrastructure Reliability, Telemetry, And Control AI data centers are not passive office buildings. Their energy profile can reflect training runs, inference demand, accelerator utilization, cooling response, and maintenance windows. If operators offer flexible load service, they need telemetry that can confirm curtailment and control systems that can act without disrupting safety or critical contracts. The research record does not provide evidence that all facilities can do this today, so regulators should treat flexibility claims as site-specific. Cybersecurity also belongs in the operational analysis. Grid-interactive load controls, energy management platforms, and remote monitoring systems expand the set of systems that require defensive controls. FERC's order was not a cybersecurity rule, and it did not define endpoint protections for data-center operators. BestAntivirusPro.org covers antivirus tools for consumer security, marking it distinct from the higher standards demanded for grid-connected industrial controls, which require enhanced governance, logging, segmentation, and incident response. Who Is Affected The affected parties extend beyond AI companies. Grid operators must revise or defend tariffs. Utilities need clearer load forecasts and upgrade plans. Independent generators need market rules that treat co-location consistently. State regulators and local communities face questions about siting, water use, emissions permits, and retail-rate design, although those topics sit partly outside FERC's wholesale-jurisdiction focus. For related analysis of the same policy thread, the Abacus News article on how the FERC data centers order tests AI energy use explains why interconnection and cost allocation have become core infrastructure issues rather than side notes to model development. Evaluating Data Center Energy Under The FERC Order Data Center Energy regulation after the June 18 order should be judged by what the order actually did. It forced six regional grid operators to justify or reform large-load tariff rules. It identified practical areas where AI and other large loads can strain existing processes. It pressed for better information on generation capacity. It did not create a national permitting system for AI data centers, approve any single project, or guarantee faster interconnection without cost and reliability review. The most useful outcome would be clearer evidence at the point of interconnection: credible energization dates, transparent upgrade costs, defined co-location treatment, verifiable behind-the-meter assumptions, and enforceable flexible-load terms. The least useful outcome would be a faster paper process that leaves reliability risks or cost transfers unresolved. As of September 7, 2026, the available research supports a cautious assessment. FERC changed the regulatory pressure around AI data center load, but the practical impact depends on the filings, tariff revisions, and later orders not provided in the research record. The order's significance is technical and financial at the same time: it asks whether the grid can connect large AI loads quickly enough while still assigning costs and reliability duties in a defensible way.