MISO's Reliability Trap: Grid Operator Drafts Rules That Codify Data-Center Privilege While Hiding the Cost

On 28 August 2026, MISO filed a proposal with the Federal Energy Regulatory Commission to impose interconnection reliability requirements on large electricity consumers, particularly data centers powering the AI boom.^[1][2] The grid operator, which runs the second-largest electricity market in the United States across 15 states and Manitoba, defined "large loads" as facilities drawing more than 50 MW at a single location and "computational loads" as large loads with at least 25 MW of demand from information technology equipment, such as servers and storage hardware.^[2][4] The filing was MISO's response to FERC's mid-June show-cause orders requiring major grid operators to establish rules for adding data centers and other large loads to the grid.^[1][2]

Read the framing carefully: MISO says it wants to "improve visibility" into large load "characteristics and behavior," to "support reliable planning and operational decision-making," and to establish "expectations proportional to demonstrated reliability risk."^[2] Translation: the grid operator is worried that data centers will suddenly trip offline, destabilizing the system, and it wants technical standards to prevent that. What the filing does NOT say is how the $billions in transmission and generation capacity built to serve these loads will be assigned to customers' bills if the data center load underperforms or moves. That silence is the mechanism.

MISO's situation mirrors the national pattern. The grid operator expects demand to accelerate from roughly 0.5% annual growth between 2009 and 2024 to 1% to 2% annually through 2044, with "higher growth rates in the near term."^[1] That acceleration is being driven by a flood of interconnection requests for data centers and advanced manufacturing, creating an unprecedented surge in large load applications.^[9] But here is the trap: reliability standards (ramp limits, ride-through duties, synchrophasor monitoring) are engineering rules, not financial rules. They tell a data center how fast it must spin up or how stable it must stay. They do NOT tell ratepayers whether a 10-year special contract or a temporary tariff will lock them into paying for a 40-year transmission asset if the customer walks away or scales back. They do NOT require the data center to sign a long minimum-take ratchet, collateral, or exit fees. They do NOT isolate the cost to a separate customer class. And they do NOT mandate that the utility study whether a flexible, curtailable interconnection service (where the data center agrees to shed load a few hours per year) would suffice instead of building firm, rate-based generation.

The Harvard Electricity Law Initiative has documented that existing tariff structures allow utilities to extract profits from the public to serve big tech. The burden of proof belongs on the utility: can this load be served with flexible grid services, or does it require new firm capacity? If new capacity is needed, who bears the risk if it is unused? MISO's filing does not address these questions. It establishes technical reliability standards without requiring simultaneous financial protections for ratepayers, which is precisely the asymmetry that lets utilities build expensive infrastructure, socialize the cost if the load underperforms, and privatize the profit if it materializes. MISO has said it will make additional proposals by a November 16 deadline; the window to demand cost-isolation protections and flexible-load alternatives in those filings is open now.

The mechanism available is straightforward: demand that MISO's next filing include a requirement that any large load above 50 MW connecting to the network must either bring its own generation and storage (BYOC), accept a separate large-load tariff with high minimum-take ratchets (80% or higher of contracted demand over the full asset life), post collateral and exit fees covering unamortized investment, or commit to curtailable/flexible grid services that allow the utility to shed that load during system peaks. That is not anti-data-center policy; it is cost-allocation discipline. A data center that can meet these terms demonstrates real commitment and removes the ratepayer subsidy. One that cannot is a phantom load, and the grid should not build for it.