Nvidia is reportedly looking to convert roughly 130 existing Brightspeed buildings across the country into small data centers, according to a report, a move that could give the AI chip company a faster route to compute capacity than building new facilities from scratch. Many of the telecom company's sites are already equipped to handle data and voice traffic and have the power infrastructure needed to run Nvidia servers. Sources cited in the report say Nvidia estimates the sites together could deliver about 1 gigawatt of compute capacity.
That capacity could house more than 2,200 Supermicro Vera Rubin NVL72 racks, each containing 72 Rubin GPUs and 36 Vera CPUs. The buildings can already supply power for the GPUs, but the report says the conversion is not plug-and-play: Nvidia would still need to install cooling systems and other supporting architecture. Those additions would require permits and approval from local governments, though the report suggests the work would likely encounter less red tape and community opposition than constructing a data center from the ground up.
Apollo Global Management has already placed significant bets on AI hardware. The asset management company is involved in a $40 billion deal with SpaceX that would give it around 360,000 Vera Rubin GPUs across 5,000 NVL72 racks. It is currently unclear who would own the AI chips installed in the Brightspeed facilities, but the proposal could help Apollo find a place for the Nvidia chips it ordered so they can be put to use as soon as possible, rather than waiting years for a new data center to be approved.
Splitting compute across multiple small sites is a setup that OpenAI and Anthropic have also been considering, particularly as larger facilities face delays and community backlash. Nvidia representatives have described such smaller sites working together as an "inference network" for running existing AI models, even though a single large, homogeneous cluster in one location is preferable for training. The report notes this matters more as AI agents, which use five to ten times more tokens than humans, grow more popular.
The idea of using existing facilities instead of building new ones has drawn renewed attention after Americans overwhelmingly voted against data centers during the midterm elections. More than 300 data center bans have been enacted across the U.S., and while they have so far only delayed three existing projects, the report says it is feared that bans and moratoriums would make new data center developments harder to start and could make it difficult for AI companies to secure the compute capacity needed to power their models in the future.
More hardware news from TechManNews.
