AI

OpenAI National Community Compact: what it means for AI buyers

Key takeaways

  • OpenAI pledges to cover its own electricity and grid costs at campuses it owns and operates
  • The commitments exclude capacity that partners build for the company
  • Enterprises should treat the compact as a signal, not a planning assumption

OpenAI has published a National Community Compact, a set of commitments to pay for the power its data centres consume and give local communities more say over construction. It arrived two days after US midterm results favoured candidates who opposed the data centre buildout supporting the AI industry. For anyone buying or planning around OpenAI's services, it is best read as a political document with some practical detail, not a capacity guarantee.

What it does well

The compact is specific about money, which is where most local resistance has focused. OpenAI says it will cover the electricity it consumes and the incremental infrastructure costs needed to serve its sites. It will also provide financial assurances before major spending begins, so that other utility customers are protected if a project is delayed, reduced or cancelled. That last point addresses a real fear: ratepayers left holding the bill for grid upgrades built for a customer that never fully arrives.

On transparency, the company commits to disclosing site-level water and electricity use, publishing construction plans, and commissioning independent noise studies before operations begin, then repeating them annually. It promises locally tailored agreements for future campuses, community input into development plans and annual public progress reports.

It also says it will stop requiring nondisclosure agreements from public officials as a condition of negotiations, keeping only narrow protections for security and commercially sensitive information. Secretive negotiations were one of the issues that fuelled opposition at the ballot box, so ending the requirement for public officials addresses one of those complaints.

Where it falls short

The biggest gap is scope. JP Perez-Etchegoyen, CTO of cybersecurity company Onapsis, pointed out that the commitments "apply only to campuses OpenAI owns and operates, not to the capacity partners build for it." Some capacity is built by partners, so the pledges do not reach all of the infrastructure OpenAI depends on.

The commitments are also voluntary. Perez-Etchegoyen noted that OpenAI's own blueprint concedes voluntary measures do not provide durable confidence. What would reduce uncertainty, he said, is the blueprint's proposals becoming law and applying to everyone who builds AI capacity, plus independent verification.

Meanwhile, the political ground is moving. Bloomberg reported that about 40% of pending US data centre capacity is planned for states where newly elected governors have promised to stop or pause construction. Bloomberg Intelligence analyst Mandeep Singh estimated that spending cuts and slower approvals could keep as much as 10 gigawatts of capacity from coming online in 2027 as scheduled. Justin Teresi, a Bloomberg Intelligence policy analyst, put it bluntly: "We've gone from the question of who is going to pass a moratorium to where can I even construct one of these things now?"

Data Center Watch's analysis of preliminary results found that 94% of elected governors support regulation or restrictions on data centre development, with 35% backing bans or moratoriums and 59% favouring regulation or conditional development. Among elected congressional candidates, 78% support regulation or restrictions.

How it compares

OpenAI is not offering a product here, so the useful comparison is between kinds of assurance an enterprise might lean on.

AssuranceBinding?Covers partner-built capacity?Independently verified?
OpenAI's compactNo, voluntaryNoAnnual public reports and independent noise studies only
OpenAI's proposed utility rules and tax-incentive termsOnly if enactedWould apply to all large customersWould add measurable commitments and recovery mechanisms

The policy side is where the weight sits. OpenAI is asking for utility rules that assign large customers the grid costs they create, easier connections for facilities that can safely reduce demand, faster transmission upgrades and expanded clean-electricity tax incentives. It also wants public tax-incentive terms and mechanisms to recover benefits when developers fall short.

Pricing and value

The source gives no service pricing, and the compact does not change what customers pay OpenAI. The costs it addresses fall on others: ratepayers, utilities and local governments. The value to an enterprise is indirect, in the chance that community goodwill keeps capacity coming online.

Who should use it

The compact is a reference point for IT leaders formulating an AI strategy, not something to adopt. Teams still formulating an AI strategy should note that the compact offers little certainty, and Perez-Etchegoyen's advice is to read it "as a signal, not a planning assumption." Those with supplier-risk or resilience duties may want to ask vendors which of their capacity is owned and which is built by partners, since the pledges only cover the former.

UK readers are not directly covered by a US compact, and the source says nothing about British sites.

Verdict

The compact is a sensible attempt to answer the complaints that led to opposition at the ballot box: who pays for power, how much water is used, and why talks happen behind closed doors. It is not a guarantee of supply. Until its proposals become law and apply to every builder, with independent checking, enterprise teams should keep their AI plans flexible and watch which states move on moratoriums.

Frequently asked questions

Does the OpenAI compact cover all of its data centres?

No. It applies only to campuses OpenAI owns and operates, not to capacity that partners build for it, according to Onapsis CTO JP Perez-Etchegoyen.

Why did OpenAI publish the compact now?

It came two days after US midterm results favoured candidates opposed to the data centre buildout, which put much of the planned construction pipeline in doubt.

More from Future Technology HQ