Oklahoma's Data Center Ratepayer Law Turns Large-Load Policy Into a Utility-Contract Model

The strongest new state-level policy signal in AI infrastructure is not coming from Washington. It is coming from Oklahoma, where the Data Center Consumer Ratepayer Protection Act of 2026 is turning a vague affordability fight into a set of utility rules that developers, regulators, and investors can actually work with. The reason the law matters is not only that it says households should not subsidize data centers. The reason it matters is that it starts specifying how that separation has to happen in practice.

House Bill 2992 defines large-load customers as new data centers, new cryptocurrency mining operations, and new facilities whose primary function is artificial-intelligence computing that add 75 megawatts or more per facility or behind a single point of interconnection after July 1, 2026. It then requires the applicable governing body to protect residential, commercial, and industrial customers from unjust rates resulting directly from service to those large loads and to allocate costs according to cost-causation principles. That threshold and that allocation language are what make this more than a symbolic anti-data-center bill.

Oklahoma's new law matters because it turns 'data centers should pay their own way' from a slogan into a utility tariff and contract structure.

The most operational part of the statute is the tariff requirement. Oklahoma says electric suppliers must establish separate terms, conditions, and tariffs for large-load customers, and those structures must include credit requirements and other measures that ensure the customer reimburses the utility for all fairly allocated costs, including costs that could remain unrecovered if the customer leaves the system or materially reduces load. The law also requires a service term of at least 10 years. That is the real shift. Oklahoma is not only telling utilities to be careful. It is telling them to build a contractual structure that directly addresses stranded-cost risk.

That requirement is already colliding with real projects. On April 30, OG&E announced it would power three new Google data centers in Muskogee and Stillwater and said the agreements would protect current customers. On May 13, after Governor Kevin Stitt signed HB 2992 into law, OG&E said Google committed to paying 100% of the costs to connect to the grid, plus its share of the costs for new generation needed to serve the sites. OG&E also said the agreement provides the framework for a new large-load tariff it plans to file with the Oklahoma Corporation Commission in the coming weeks. That sequence matters because it shows the law is arriving alongside a real utility-customer template rather than as a disconnected political statement.

The KOSU and StateImpact Oklahoma reporting sharpens the timing. Their May 15 report says Public Service Company of Oklahoma has already filed a large-load tariff with the Oklahoma Corporation Commission, while OG&E is preparing its own filing. That makes Oklahoma one of the clearest near-term tests of whether states can convert generic ratepayer-protection rhetoric into commission-ready tariff design before large-load growth disappears into broader rate structures.

For The Grid Report, the original angle is that Oklahoma is operationalizing the national large-load debate. The site has already covered the broad question of who pays for AI data-center grid upgrades. Oklahoma is different because it turns that abstract question into a package of definitions, credit standards, tariff separation, minimum service terms, and filing obligations. In other words, it makes the AI power story look less like a one-off political fight and more like a repeatable utility-contract model.

The investor read is also more specific than a simple 'utilities benefit from load growth' story. HB 2992 is about whether large-load growth can be underwritten without leaving ordinary customers exposed if a customer cancels, delays, downsizes, or consumes less power than expected. That is the key risk embedded in the AI data-center buildout: utilities may need to plan generation, transmission, substations, and procurement around customers whose load ramps are large, politically sensitive, and still uncertain. A separate tariff with credit support and a 10-year minimum term is one way to make that risk visible before the capital is spent.

That has implications well beyond one state. Operators now have a clearer signal about what utilities may ask for when hyperscale campuses want speed and firm power. Investors have a better read on how utilities and commissions may try to ring-fence stranded-cost risk before approving major load additions. And policymakers in other growth states now have a recent example of how to move from anti-subsidy language into utility mechanics. If Oklahoma's filings hold up, HB 2992 may matter less as a local law than as a template for the next wave of data-center tariff battles.

Oklahoma House of Representatives, House Advances Rep. Boles Bill to Protect Ratepayers from Data Center Energy Costs, March 23, 2026: https://www.okhouse.gov/posts/news-20260323_3

Oklahoma Legislature, HB 2992 floor version, accessed May 18, 2026: https://www.oklegislature.gov/cf_pdf/2025-26%20FLR/HFLR/HB2992%20HFLR.PDF

OG&E, “OG&E Announces Landmark Contract with Google,” April 30, 2026: https://www.oge.com/web/portal/-/200-press-release

OG&E, “OG&E welcomes enactment of House Bill 2992,” May 13, 2026: https://www.oge.com/web/portal/-/201-press-release

KOSU / StateImpact Oklahoma, “Stitt signs bill to prevent higher utility costs from data centers into law,” May 15, 2026: https://www.kosu.org/data-center-utility-cost-bill-signed