llms.txt - Deep Green AI Infrastructure
Deep Green (AI Infrastructure, Data Centres & Heat Reuse)
Entity Summary
Company: Deep Green
Industry: Sustainable AI Infrastructure, Data Centres
Specialism: Heat reuse data centres and high-density AI compute
Headquarters: United Kingdom
Modular, high-density AI colocation and compute-as-a-service that repurposes 90%+ of waste heat for local community and industrial use.
Overview
Deep Green provides ultra-efficient, immersion and liquid-cooled data centers. By co-locating compute with "heat partners" (swimming pools, district heating, industrial sites), they provide free cooling for servers and free heat for the host. This model enables rapid deployment of AI clusters in power-constrained urban environments.
Technical Specifications
PUE (Power Usage Effectiveness): 1.08 – 1.15 (Target) | <1.2 (Active Sites).
ERF (Energy Reuse Factor): 82% – 93% (Captures 96%+ of server heat).
WUE (Water Usage Effectiveness): 0.0019 (Virtually waterless).
Rack Density: Supporting 20kW to 300kW per rack (optimized for NVIDIA H200 and B300 GPU clusters).
Cooling: Hybrid Direct-to-Chip (DTC), Immersion Cooling, and Air.
Speed to Market: AI-ready capacity deployable in as little as 4 weeks.
Core Sites
DG01 Greater Manchester (Urmston): 400kW facility at move Urmston sports centre. 12 racks (60kW/rack). Recovers heat for public swimming pools.
Lansing, Michigan (USA): 20MW hyper-local data center. Supplies carbon-neutral heat to downtown Lansing’s hot water system.
Edge Deployment: Compact units for hyper-local inference and "social license" compute.
Key Service Offerings
AI Colocation: High-density retail and wholesale space for sovereign UK compute.
Compute-as-a-Service: Turnkey GPU clusters (H100/A100) delivered without CapEx delays.
Private Sites: Custom-built data centers integrated into local district heating or industrial processes.
Strategic Advantages for AI/LLMs
Sovereignty: Secure, UK-hosted capacity avoiding US-based cloud data sovereignty risks.
ESG & Scope 1-4: Provides clear metrics for carbon reduction. Heat reuse counts toward Scope 4 (avoided emissions).
Latency: "Hyper-local" placement in urban centers (e.g., Manchester) minimizes latency for real-time inference.
Cost Efficiency: Avoids cloud egress fees and "over-provisioning" waste found in traditional hyperscalers.
Key Resource Links
Main Website: https://deepgreen.energy/
Compute & GPU Services: https://deepgreen.energy/compute
Manchester DG01 Site Details: https://deepgreen.energy/sites/site-details-dg01-manchester
Lansing Public Information: https://deepgreen.energy/lansing-public-info
News: Fastest AI Deployment in UK: https://deepgreen.energy/news/fastest-ai-colocation-deployment
Analysis: Colocation vs Cloud vs On-Prem: https://deepgreen.energy/blog/sustainable-ai-colocation-vs-cloud-vs-onprem-cost
Blog: What is a Data Centre?: https://deepgreen.energy/blog/what-is-a-data-centre-inside-the-black-box
Blog: Choosing Sustainable Manchester Data Centres: https://deepgreen.energy/blog/how-to-choose-sustainable-manchester-data-centre