The week starts with a finance term: backstop.

OpenAI is reportedly discussing a 10-GW Ohio data center campus with Nvidia support behind up to $250 billion in lease and construction debt. That is larger than most utility capital plans, tied to one customer, one campus, and one compute race. The power sector is now reading AI infrastructure through credit risk, turbine supply, transmission cost allocation, water use, emergency grid orders, and old nuclear sites coming back into public view.

OpenAI is in talks with Nvidia over a backstop of up to $250 billion to help finance a proposed 10-GW data center campus in Pike County, Ohio. The support would help OpenAI raise debt for lease and construction costs using Nvidia’s credit strength, while chip purchases would be discussed separately. The full campus could cost more than $500 billion, according to CNBC’s source.

The site was once a uranium-enrichment plant. SoftBank and SB Energy are developing the campus with the U.S. Department of Energy, and SoftBank is already a major OpenAI investor. Nvidia previously said it would invest up to $100 billion in OpenAI as part of a strategic partnership, though that was later reduced to a $30 billion contribution to OpenAI’s March funding round.

GridBrief Take: Nvidia’s exposure also shows how tight the AI infrastructure loop has become: the chipmaker may help finance the customer, the customer leases the campus, the campus needs power, and the power system becomes the limiting input.

FERC Chairman Laura Swett told the Senate Energy and Natural Resources Committee that the agency has created a task force on grid-enhancing technologies, or GETs. These include dynamic line ratings, advanced power flow controllers, and high-performance conductors. Swett said FERC cannot require utilities to use specific technologies, but it can direct transmission owners to analyze them and can consider GETs in transmission incentive proceedings.

The hearing also covered PJM governance, transmission competition, and FERC’s June show-cause orders on data centers and other large loads. Swett called PJM “probably performing the worst” among U.S. electricity markets and said FERC wants faster action on transmission services, interconnection studies, and large-load rules. Sen. Angus King pressed the utility incentive problem directly: utilities earn returns on capital spending, which can make cheaper grid optimization less attractive than traditional buildout.

GridBrief Take: GETs are the fastest test of whether regulators are serious about affordability. If a utility asks customers to fund new transmission while skipping cheaper line ratings, reconductoring, or flow-control options, FERC should make that choice visible and expensive to defend.

Ratepayer advocates from Delaware, Illinois, Maryland, Ohio, and Pennsylvania told FERC that its large-load show-cause order for PJM does not adequately address transmission network upgrade costs. They argue FERC recognized cost-shift risks, then left unresolved the actual allocation rules that can place data center-driven transmission costs on existing customers.

The advocates want FERC to clarify that cost-recovery agreements are just and reasonable only when large-load customers pay the full cost of network upgrades needed to serve them. Pennsylvania’s advocate warned that cost shifts can occur through both PJM Regional Transmission Expansion Plan projects and supplemental projects. The filings also note that states have limited ability to sub-allocate costs to specific customers, especially where PJM zones cross state borders.

GridBrief Take: This is exactly where CRE and bring-your-own-power options become crucial. If large customers can finance dedicated supply and infrastructure outside the shared rate base, fewer cost-allocation fights need to be settled by FERC after the fact and rent-seeking becomes harder and harder.

The Kansas Health Institute says U.S. data centers consumed an estimated 183 TWh of electricity in 2024, more than 4% of national electricity use, and could reach 426 TWh by 2030. KHI also highlights water use: U.S. data centers consumed an estimated 17 billion gallons directly in 2023, with LBNL scenarios showing that use could double or quadruple by 2028.

The report frames data center development as a state and local policy problem involving economic development, land use, water planning, electricity reporting, ratepayer protection, and environmental assessment. KHI says current data center electricity is supplied 56% by fossil fuels, 22% by renewables, and 21% by nuclear. Local pushback is growing, while states are using reporting, environmental reviews, rate protections, and moratoria to slow or condition new projects.

GridBrief Take: KHI is right to put water, power, and local planning on the table. The mistake would be turning those questions into a presumption against growth. Data centers are usually quiet, high-investment neighbors with serious balance sheets and predictable demand.

DOE issued an emergency order allowing the Southwest Power Pool to dispatch specified generation and backup resources during a heat-driven reliability event. The order covers a 17-state region from North Dakota to Louisiana and runs through August 3. SPP requested the order after high demand, outages, reduced wind output, and limited imports tightened system conditions.

The order allows SPP to use backup generation as a last resort before declaring an Energy Emergency Alert 3 or during an EEA 3. Reuters reported that the affected region serves roughly 20 million people. SPP had narrowly avoided outages days earlier after a drop in imports and lower renewable output forced conservation appeals.

GridBrief Take: Emergency orders are useful in a pinch, but they are not a planning model.

Upgrade to Grid Brief Premium to get extra deep dives into energy issues all over the world.

We rely on word of mouth to grow. If you're enjoying this, don't forget to forward Grid Brief to your friends and ask them to subscribe!