Nvidia Eyes Old Telco Buildings for Tiny Data Centers to Bypass Community Backlash
Nvidia Eyes Old Telco Buildings for Tiny Data Centers to Bypass Community Backlash
Nvidia is reportedly in talks to repurpose over 130 telecommunications buildings into compact data centers to bypass community opposition and lengthy approval processes for new infrastructure. The proposed sites could collectively deliver around one gigawatt of compute capacity using advanced hardware.
As demand for artificial intelligence infrastructure continues to surge, technology giants are seeking alternative strategies to secure necessary compute capacity. Nvidia has pitched an innovative proposal to convert existing telecommunications facilities into distributed tiny data centers. By utilizing pre-existing buildings with established power connections, companies hope to accelerate deployment while avoiding the severe community resistance and regulatory hurdles that frequently plague massive greenfield developments. This approach reflects a broader industry shift toward decentralized architectures designed to support growing inference workloads.
Repurposing Legacy Infrastructure
Internet and telecommunications provider Brightspeed is reportedly engaged in discussions with Nvidia regarding the conversion of several properties into compact data centers. Nvidia initially pitched the concept to executives at Apollo Global Management, the parent company of Brightspeed, leading to a formal request for written operational plans.
Following the submission, Nvidia has actively requested additional technical data, detailed information, and on-site visits to evaluate Brightspeed facilities. These properties already possess vital power infrastructure designed to handle telecommunications traffic, making them uniquely positioned to house modern high-density hardware configurations.
Brightspeed Discussions
Nvidia pitched the infrastructure conversion idea to Apollo Global Management, prompting formal requests for written operational plans.
Infrastructure Evaluation
The AI tech giant requested further site visits and technical data to evaluate existing telecommunications properties.
Compute Potential and Hardware
The targeted telecommunications buildings could collectively deliver roughly one gigawatt of compute capacity. This power footprint is estimated to accommodate more than 2,200 Supermicro Vera Rubin NVL72 racks, each packed with advanced processing units.
Deploying hardware in these locations is far from a plug-and-play operation. Specialized cooling systems and supporting architectural upgrades must be installed, requiring local government permits and regulatory approval despite facing less community opposition than entirely new developments.
Overcoming Regulatory Bottlenecks
The push to utilize standing facilities gains urgency as public opposition to large data centers intensifies across the United States. Recent midterm elections saw voters reject numerous proposed projects, leading to widespread moratoriums and bans on new greenfield construction.
Developing distributed small data centers offers a practical workaround to avoid years of administrative delays. Similar network topologies are also under active consideration by leading artificial intelligence labs seeking alternative deployment pathways.
Voter Resistance
Americans overwhelmingly voted against data center developments during recent midterm elections, enacting numerous regional bans.
Regulatory Hurdles
Distributed sites face significantly less red tape compared to greenfield data center developments that trigger severe community pushback.
Building an Inference Network
While homogeneous clusters remain preferable for initial AI model training, industry experts describe smaller interconnected sites as an inference network. These distributed nodes are well-suited for running existing models efficiently as user demands shift toward resource-intensive applications.
As modern artificial intelligence agents consume significantly more tokens than human users, optimizing infrastructure for inference becomes critical. Utilizing diverse geographic nodes helps maintain operational flexibility while mitigating power constraints on primary regional grids.
Why It Matters
The exploration of legacy telecommunications buildings for artificial intelligence compute highlights a pragmatic evolution in infrastructure strategy. As public pushback and regulatory hurdles stall greenfield data center projects across the United States, technology firms must innovate to secure necessary power and space. Repurposing existing real estate with established utility connections reduces deployment friction and aligns with the operational demands of modern inference workloads. While significant engineering efforts like advanced cooling integration remain necessary, this decentralized model provides a viable roadmap for scaling artificial intelligence capacity sustainably.
Nvidia wants to repurpose over 130 old telco buildings as tiny data centers, report claims (tomshardware.com) · 'Families need reliable electricity': OpenAI pledges to try and provide 'a national effort to support affordable, reliable electricity for families and businesses' living near its data centers (techradar.com)
Reporting reviewed through November 6, 2026.
