Nvidia pitches Apollo on converting 130+ Brightspeed telecom buildings into small AI data centers to sidestep data-center moratoriums (Nov 5)
What happened
Nvidia has approached private equity firm Apollo Global Management about converting more than 130 telecom buildings owned by Apollo-backed broadband provider Brightspeed into a network of small AI data centers spanning as many as 20 U.S. states, The Information reported on Thursday, Nov 5. Nvidia submitted a written proposal to Brightspeed on Wednesday and has requested additional data and site visits. The sites together hold roughly 1 gigawatt of available power, which Nvidia estimates could support tens of billions of dollars' worth of its upcoming Vera Rubin servers, with retrofits potentially beginning in Q1 2027. Nvidia describes the proposed sites as an 'inference network' for running already-trained models rather than large-scale training clusters, with potential customers including SpaceX, Anthropic and OpenAI.
Why it matters & what’s next are part of Pro, along with the audio summary.
Start your free 14-day trialWho’s affected & sources
Who’s affected
- Nvidia (NVDA): a workaround that keeps its chip deployments growing even if new mega-campuses are blocked
- Apollo Global Management (APO): a potential new revenue stream for a distressed portfolio asset and a deepening tie to Nvidia's hardware cycle; hyperscalers and neoclouds (MSFT, AMZN, GOOGL, META, CoreWeave, Nebius, IREN): the distributed-inference model could ease their own siting constraints, though Microsoft separately declined to extend a major data-center agreement with Nvidia-backed startup Nscale; telecom carriers (AT&T, Verizon, Lumen): AT&T alone holds roughly 5,000 central offices, suggesting the retrofit playbook could extend far beyond Brightspeed's 130 buildings.