OpenAI renames its data centers as “compute campuses”

OpenAI introduces a new term, “compute campuses,” to refer to its data centers. Behind this vocabulary change, the company mainly seeks to ease criticism regarding electricity, local networks, and environmental footprint.

OpenAI has begun renaming its data centers as "compute campuses" in an announcement published Monday, as local opposition to AI infrastructure rises in the United States. The first site involved, being prepared in Effingham County, Georgia, is part of a program presented as a response to tensions on power grids. The change in terminology does not alter power requirements, but it says a lot about the political pressure now surrounding AI. And vocabulary has clearly become a project in its own right.

A response to criticism about electricity and grids

OpenAI's "Grid Protectors Program" aims to reduce the pressure its future facilities place on power grids, especially during demand peaks linked to cold or heat waves. According to the company, these sites will be able to reduce their consumption during critical periods through "demand-response" mechanisms, that is, temporary adjustment of electricity demand based on grid status. OpenAI also promises to cover the costs it generates and help communities to "keep the lights on."

In its communication, the company emphasizes the flexibility of these new sites, contrasting them with data centers "built only a few years ago." It also mentions adding on-site batteries and auxiliary power controls to smooth consumption variations. The message is clear: to present these infrastructures as an asset to the grid, rather than just an energy sink.

Why this name change is happening now

The timing is no accident. According to Axios, OpenAI was already preparing a communication and lobbying offensive after the U.S. midterm elections, in a potentially more hostile political context. The term "compute campus" appears for the first time in the company's communication, after searches on its site found no prior usage.

On the critics' side, the move mostly looks like lexical varnish. Holly Elmore, founder of PauseAI US, believes the new name "may fool listeners once or twice but it’s not the name ‘data center’ that is unpopular — it’s the giant, polluting construction projects." Abre’ Conner, from the NAACP's Center for Environmental and Climate Justice, calls it "an empty solution to an existing issue." In other words: you can repaint the facade, but the generator noise remains the same.

A real market issue, not just communication

At its core, OpenAI is trying to address a problem that goes beyond its case. The capacity of data centers in the United States could exceed 100 gigawatts by 2030, according to estimates cited by sources, driven by the computing needs of generative models. Meanwhile, a Gallup poll relayed by CNET indicates that 70% of American adults oppose building data centers near their homes. This opposition crosses partisan lines and turns the siting of these sites into a municipal, energy, and electoral issue.

OpenAI moves forward with a concrete example in Georgia: an agreement with Georgia Power could free up to 1,000 megawatts of flexible demand during grid emergencies, equivalent to powering about 700,000 homes. The company also says it wants to fully finance its baseline electricity and the dedicated infrastructure needed for its sites. Enough to calm some local officials, but not necessarily residents or associations, who demand broader commitments on water, emissions, and transparency.

OpenAI facing Anthropic and the rest of the sector

This strategy also recalls the competition among major AI labs. Anthropic promised this year to cover 100% of the costs of connecting its data centers to grids, including those often passed on to consumers, and to invest billions of dollars in community projects. OpenAI seems to be seeking its own line of defense, between discourse on local responsibility and repositioning its public image.

The central question remains: is changing the name enough to change perception? For now, sources agree on a simple point. OpenAI's "compute campuses" look very much like data centers. They consume enormous amounts, are located in already strained territories, and force AI companies to explain, with figures, why their needs should be accepted by the public.

By the numbers

  • 1,000 megawatts — flexible demand announced in Georgia, according to OpenAI.
  • 700,000 homes — equivalent capacity mentioned for the Georgia site.
  • 100 gigawatts — projected capacity of data centers in the United States by 2030.
  • 70% — share of American adults opposed to a data center near their home, according to Gallup.
  • 100% — connection costs covered by Anthropic, according to cited sources.

To read