What's Happening?
Perimeter Compute, a startup emerging from stealth, plans to leverage excess energy capacity in commercial office buildings across major U.S. cities to establish edge AI data centers. The company, part of Montauk Capital's portfolio, aims to install AI chips
in buildings with significant spare capacity, addressing current data center construction bottlenecks. CEO David Hall highlights that many buildings have reduced power consumption significantly due to energy conservation efforts, yet maintain infrastructure designed for peak loads. By utilizing this excess capacity, Perimeter intends to support AI inference tasks, which require less energy than training large language models. The company is currently raising capital to purchase equipment, with costs estimated at $35 million per megawatt. Perimeter is also engaging with local utilities to manage increased loads and ensure grid stability.
Why It's Important?
This initiative by Perimeter Compute could significantly impact the AI and data center industries by providing a more efficient and distributed approach to processing AI workloads. By utilizing existing infrastructure in urban centers, the company can offer faster processing capabilities, which is crucial for applications like chatbots and other AI programs. This approach not only addresses the current challenges faced by data centers, such as equipment shortages and power grid constraints, but also offers a sustainable solution by repurposing underutilized resources. The potential partnerships with tech companies and building owners could lead to new revenue streams and operational efficiencies, benefiting both the tech industry and real estate sectors.
What's Next?
Perimeter Compute is in the process of raising funds and negotiating contracts with building owners and tech companies. The success of these negotiations will determine the scalability of their model. The company is also working with local utilities to manage the expected increase in power demand, which could influence future energy policies and infrastructure planning. As the edge computing market grows, Perimeter's approach may set a precedent for similar initiatives, potentially leading to widespread adoption of distributed AI processing solutions in urban environments.
