Nvidia has submitted a proposal to turn more than 130 telecommunications buildings owned by Brightspeed into small AI data centers across the United States, according to The Information. The idea emerges as a way around growing opposition to large data-center projects — but for now it is only a proposal: there is no signed deal, no confirmed customer list and no approved conversion program.
What was reported
According to The Information, citing two people with direct knowledge of the matter, Nvidia approached Apollo Global Management, which owns Brightspeed, with a plan to build a network of small data centers in up to 20 US states. Apollo asked for the proposal in writing, and Nvidia delivered it, requesting technical information and visits to several sites.
Nvidia's own estimate, as reported, is that the buildings add up to roughly 1 gigawatt of available power — enough to run systems with Vera Rubin chips that, according to The Information, could be worth tens of billions of dollars. Locations mentioned include Dothan, Alabama, Killeen, Texas, and Mansfield, Ohio.
Brightspeed was created in 2022, after Apollo funds bought part of Lumen Technologies' local telephone and broadband operations for US$ 7.5 billion. People familiar with the matter told The Information the company was reportedly nearing bankruptcy — which is not confirmation of a bankruptcy filing.
Why old telephone buildings
The buildings already have power, fiber and approved zoning, which could speed up deployment compared with building a new campus. The proposal comes amid strong backlash against large data centers: at least 40 local laws to curb or scrutinize construction of such facilities were passed in US local elections, according to The National, and winning gubernatorial candidates, including in Texas, called for moratoriums on new projects.
The proposed network would be for inference — running already-trained AI models rather than training new ones. Nvidia has reportedly discussed starting retrofits in early 2027, with Vera Rubin servers and Groq technology, but there is no confirmed schedule. Potential customers could include SpaceX, Anthropic and OpenAI, although it is unclear whether Nvidia or Apollo has talked to those companies about the project.
There are practical obstacles: older telecom buildings may need specialized cooling systems and electrical upgrades. Zac Smith, chief executive of AI networking company Datum, told The Information that such properties can unlock underused electricity, but warned they can sometimes cost more to operate because of aging electrical equipment.
The Nvidia context
The move connects to two themes Geekiki has already covered: pressure over AI costs, which dragged Nvidia shares down and pushed the Nasdaq into a correction in early November, and Apollo's relationship with Nvidia — the investment firm recently agreed to buy tens of billions of dollars' worth of Vera Rubin chips to lease to SpaceX. None of that means SpaceX will use Brightspeed's buildings: the report makes clear that has not been established.
In short, the proposal is a behind-the-scenes report, not an official announcement. If it moves forward, it could show a new way to expand AI computing capacity without relying on mega-projects — but how many buildings are suitable, what the retrofits would cost and how much power they would actually deliver remain open questions.
Image credit: NVIDIA logo, via Wikimedia Commons.
Sources
- TechTimes — Nvidia Is Looking at Old Telephone Buildings for AI Data Centres (November 6, 2026)
- FourWeekMBA — Nvidia Said to Pitch Deal on Old Phone Buildings With 1 GW
- Geekiki — Nvidia falls and Nasdaq enters correction amid AI cost fears
- Geekiki — Update: SpaceX Apollo details US$ 40 billion GPU lease
Transparency: this article was produced with the assistance of artificial intelligence, cross-checked against public posts on X and internet sources. Information attributed to The Information is a report based on anonymous sources, not a fact confirmed by the companies. Check the original sources in the links above before sharing.
By Muse
