OpenAI to Press On With Compute Build-Out as Voters Curb Data Centres

OpenAI will keep investing in artificial intelligence infrastructure and continue raising capital, its chief financial officer said, after voters turned against data-centre projects in this week's midterm elections. Speaking to CNBC at the Citi TMT Leadership Summit at Pebble Beach in California on Friday, Sarah Friar said the company must "keep investing to meet that customer demand" and that the results had not dramatically changed anything for OpenAI. The position bears directly on the industry's central constraint: developers depend on data centres to train models and serve customers, and the midterms have turned the siting of that compute into a local political question. Out of 41 local ballot measures across 10 states, voters backed restrictions or rejected pro-development measures for all but one, according to a Ballotpedia analysis and additional results gathered by CNBC.

Inside the company, chief executive Sam Altman took questions from staff at an internal Q&A on Wednesday, The Wall Street Journal reported, in an account carried by The Times of India. According to that report, Altman acknowledged concern about how much water the facilities use and said he understood why people might not want to live near one, at one point musing that OpenAI should perhaps look at building data centres in the desert. Senior leaders discussed whether OpenAI would need to reset expectations for its computing capacity next year, and staff working on site development were swapping notes on newly elected Democratic politicians in key states. On Thursday the company published a National Community Compact covering future sites it owns and operates, which it now prefers to call compute campuses: it said it will not ask public officials to sign non-disclosure agreements, barring narrow protections for security information and trade secrets, will pay the full cost of the electricity it uses, publish site-level water and power use and release independent noise studies before operations begin. At a WSJ tech conference, OpenAI president Greg Brockman said the company needed to do a better job of explaining to local communities why data centres can benefit them; Meta president Dina Powell McCormick told the same event that building new facilities in the United States is going to be harder and that technology firms are increasingly weighing sites outside the country.

Candidates who promised to freeze new data-centre development won governor's races in Texas, Georgia, Ohio and Michigan. Analyst Dan Ives has said 25% to 30% of upcoming data-centre capacity will come from Texas, where governor-elect Gina Hinojosa has said projects already permitted will be allowed to finish construction, covering most of what is due in 2027 and likely into 2028; her tougher rules target newly permitted sites, which she wants to pay taxes, avoid pushing up power bills, protect local water and use union labour. On the All-In podcast, investor Brad Gerstner said as much as 35% of US data-centre construction planned for next year is now at risk. Investors have turned sceptical on the AI trade amid higher borrowing costs, concerns about soaring debt and public backlash towards the industry, and the Nasdaq on Thursday suffered its steepest drop in 19 months, led by chipmakers and neoclouds, before rebounding somewhat on Friday.

OpenAI confidentially filed its prospectus with the Securities and Exchange Commission in June, and Friar said on Friday the company would continue to look into a 2027 debut. She said OpenAI is in talks to raise funding at a $1.25 billion pre-money valuation and that the round is ongoing. Rival Anthropic, which is preparing a public listing and has flagged moratoriums as a risk to its compute plans, aims to reach the Nasdaq in mid-November; its prospectus reports revenue of $33.2 billion in the first nine months of the year, up roughly 1,200%, and purchase commitments and lease agreements with aggregate committed spend of about $895.4 billion, substantially all related to compute and infrastructure. Friar's reading of demand was blunt: "The demand for AI is real. I don't think anyone's arguing with that, or if they are, I think it's a losing argument."