After midterm rout, OpenAI tries to win over communities it needs for AI expansion

OpenAI has published a National Community Compact and a Policy Blueprint setting out voluntary commitments for future data center sites, issued two days after US midterm elections in which voters rejected new data centers in 40 of 41 local ballot measures. The company is calling on lawmakers to make the rules binding across the industry, not just for itself. The timing suggests a response to organized local resistance to AI infrastructure buildout; OpenAI's policy chief Chris Lehane has been pitching the commitments to skeptical residents.
The stated scope is limited. The pledges cover future sites OpenAI owns and operates directly. A large share of the company's data center infrastructure is built and held by partners such as Oracle, so those facilities fall outside the compact. That gap matters because the voters who blocked ballot measures were reacting to construction near them, whoever the eventual operator is.
At its own sites, OpenAI says it will stop requiring nondisclosure agreements during negotiations with local authorities, with carve-outs for trade secrets, security concerns and competitively sensitive information. It also promises to cover the full cost of electricity and grid upgrades, throttle consumption during peak hours, and disclose construction plans, noise studies and water usage data. In practice, those are the three pressure points that typically drive local opposition: strain on the power grid, noise from cooling and generators, and water consumption for cooling. Powering and upgrading the grid at company expense shifts a cost that often lands on ratepayers, and publishing water and noise data gives residents something concrete to check against promises.
Lehane's argument leans on economic benefits. He told Reuters the facilities would not be a black box, because regular hardware upgrades would generate permanent jobs rather than one-off construction work. He cited 2,500 positions for Ohio. OpenAI also points to local payments such as $4.8 million for school lunches in Georgia. Set against planned project costs of more than $30 billion in that state, such sums read as largely symbolic next to the capital being deployed.
The political backdrop is sharp. Beyond the ballot defeats, Democrats won not only Congress but governor's seats in infrastructure-heavy states including Texas, Georgia and Ohio. Texas' new governor, Gina Hinojosa, has called for an immediate moratorium on new facilities. Research group Data Center Watch reported that 45 projects worth $88 billion were already stalled by local opposition in the second quarter of 2026. Lehane acknowledged that OpenAI projects could face delays as well.
Markets split along size. Specialized data center operators fell on fears of tighter permitting: Hut 8 dropped 23 percent, IREN 11 percent and Core Scientific 10 percent. Cash-rich hyperscalers moved the other way, with Microsoft up 6.7 percent, Amazon 6.1 percent and Alphabet 3.2 percent. The reading among investors appears to be that a slower buildout curbs runaway infrastructure spending and favors companies that already run large data centers. This suggests a consolidation effect: permitting friction acts as a barrier to entry that incumbents can absorb and smaller operators cannot.
One of the election's most visible losers was Elon Musk. According to Politico, he put a record $446 million into Republican candidates, more than $200 million of it in Texas, and nearly all the politicians he backed lost. That leaves his own data center expansion plans in Texas on uncertain footing. SpaceX shares fell 3.9 percent and Tesla 3.3 percent. Musk responded on X by pointing to orbital infrastructure plans, writing that data centers in space look appealing right now.
For developers and product teams, the near-term risk is schedule slip. Training runs and inference capacity depend on facilities clearing local approval, and a stalled project can push back model availability or raise API pricing if capacity gets tight. Inference, the serving of trained models to users, is especially sensitive to power costs and peak-hour throttling, which the compact explicitly allows at OpenAI's own sites. A likely trade-off is that voluntary commitments buy goodwill but do little to bind competitors, which is presumably why OpenAI wants them legislated. Whether partner-built capacity gets folded into similar terms is the open question, and it is the one that determines how much of the buildout the compact actually covers.
Why it matters: teams planning around OpenAI capacity should treat construction timelines as less predictable, since the company itself acknowledges possible delays. Smaller infrastructure providers face the sharpest permitting risk, while hyperscalers stand to gain share if the buildout slows. If the commitments become binding law, operators across the industry would inherit grid-cost, disclosure and peak-throttling obligations that change how data center economics are modeled.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
OpenAI is responding to the clear vote by US voters against new data centers with voluntary commitments for future sites. The pledges only apply to facilities OpenAI operates itself, though, and a large share of its infrastructure is built by partners like Oracle. Voters rejected new data centers in 40 out of 41 local ballot measures. The article After midterm rout, OpenAI tries to win over communities it needs for AI expansion appeared first on The Decoder .