OpenAI plans to continue expanding its US data-centre footprint despite political pressure and questions about the cost of large AI infrastructure projects, chief financial officer Sarah Friar told CNBC, according to Reuters. The company is also changing how it discusses proposed sites with local officials. Friar said OpenAI would stop using non-disclosure agreements with public officials during site discussions and would share construction plans and potential community impacts with residents.
The change comes as data-centre projects have become a visible local and national issue. Large facilities require substantial electricity, cooling capacity, land, fibre connectivity and supporting infrastructure. Communities have raised questions about power demand, water use, noise, construction traffic and whether new facilities will raise costs for existing electricity customers. In some areas, opposition has moved from local planning meetings into election campaigns. The debate is no longer limited to whether AI infrastructure should be built, but who bears the costs and who receives the economic benefits.
Friar’s comments followed a period of market volatility in which investors reassessed the pace and economics of AI infrastructure spending. Reuters reported that concerns about tighter scrutiny after the US midterm elections had contributed to investor anxiety. OpenAI has argued that enterprise demand remains strong, with Friar citing quarter-over-quarter business growth and a doubling of enterprise demand in the third quarter. Those figures are company statements rather than independent measures of long-term profitability, and they do not by themselves settle whether planned capacity will earn an adequate return.
OpenAI is changing how it negotiates with host communities
Bloomberg’s report on the company’s revised community-engagement commitments. Ending the use of NDAs with public officials is a notable process change because it addresses a criticism often made by residents: that major infrastructure decisions are negotiated without enough public visibility. Sharing plans and potential impacts earlier could give local governments and communities more time to evaluate energy supply, water requirements, road access and emergency-service needs. It may also make project development more complex, because companies will have to answer questions before every detail is final. The practical value will depend on how complete and accessible the disclosures become.
Data-centre development is not a single construction decision. Projects depend on power purchase agreements, grid interconnection, transmission upgrades, backup generation, cooling systems, land approvals and equipment providey. Even when a company announces a large investment, the schedule can change if grid capacity or permitting is delayed. Communities may also want enforceable commitments on local hiring, tax contributions, water recycling and the use of low-carbon electricity. A public engagement process can expose these dependencies earlier, but it cannot remove the physical constraints of building at scale.
For OpenAI and other AI companies, infrastructure planning is becoming a test of operational credibility. Customers want assurance that computing capacity will be available, while investors want to understand capital requirements and utilisation. Local authorities want evidence that a project will not impose disproportionate costs on households or businesses. A credible proposal needs to connect the expected computing demand with a realistic energy plan, a financing structure and measurable community benefits. Broad promises about AI growth are unlikely to answer questions about a specific substation, water source or local road network.
Grid access and local approvals will shape expansion schedules
The policy debate also has implications for the geography of AI development. Regions with available power, supportive permitting and established fibre networks may attract projects, while constrained grids could push investment elsewhere. That can create uneven local outcomes: some areas may gain construction work and tax revenue, while others experience higher infrastructure pressure. National policy can help coordinate transmission and energy investment, but it must also address how costs are allocated. OpenAI’s stated transparency change is one company response to this conflict, not a substitute for consistent public rules.
The next measure will be implementation. Communities will be able to judge whether OpenAI provides usable project information, responds to concerns and follows through on commitments. Investors will watch whether the company can secure power and complete facilities on schedule. Customers will watch whether capacity arrives at the pace promised. The company’s position is that expansion remains necessary, but the conditions for building have changed. AI infrastructure operators increasingly need a social licence alongside land, power and capital, and transparent engagement is becoming part of the providey plan.