Metrobloks —
Next-Generation AI & High-Density Colocation

Metrobloks is purpose-built for the next generation of computing workloads — particularly AI and high-performance computing that exceeds the power and cooling limits of traditional data centers. Metrobloks supports 150 KW or more per rack through liquid-to-chip cooling, compared to the 5–15 KW per rack typical of air-cooled facilities. The open floor design provides maximum flexibility for customers with unique infrastructure requirements. Tax abatement in select markets further reduces total cost of ownership compared to standard colocation pricing.

Liquid-to-chip cooling routes coolant directly to the processors — GPUs, CPUs, and other compute components — rather than cooling the surrounding air. This allows dramatically higher heat dissipation per rack, enabling 150 KW or more per rack compared to traditional air cooling limits of roughly 15–30 KW per rack even with advanced air cooling. For AI and machine learning workloads running dense GPU clusters, liquid cooling is not optional — it is required to achieve the compute density that modern AI training and inference demand.

Wholesale colocation provides large blocks of power, space, and cooling to a single customer — typically measured in megawatts rather than individual racks or cabinets. It is designed for hyperscale operators, large cloud providers, enterprises running private cloud infrastructure, and AI companies building large-scale GPU clusters. Wholesale customers typically have significant power requirements (1 MW or more) and benefit from the economics of bulk capacity and long-term agreements.

Retail colocation provides per-rack and per-cabinet space in a shared data center environment — designed for organizations that need data center space without committing to bulk power or large footprints. Metrobloks retail colocation includes redundant power, cooling, physical security, and 24/7 remote hands support. It is appropriate for enterprises, mid-market companies, and organizations deploying traditional workloads alongside high-density AI infrastructure in the same facility.

Yes. Metrobloks supports both air cooling and liquid-to-chip cooling within the same facility. This is important for customers transitioning from traditional air-cooled infrastructure to high-density AI deployments — existing workloads can remain in air-cooled racks while new GPU clusters are deployed in liquid-cooled high-density zones. Customers are not forced to migrate everything immediately.

Metrobloks has secured tax abatement agreements in certain markets, which can meaningfully reduce the total cost of operating large-scale data center infrastructure. Tax abatement typically applies to property taxes and equipment taxes and is passed through to qualifying customers. The specific markets and abatement terms vary — during the evaluation process, Fibi can help identify which Metrobloks locations offer abatement that would apply to your deployment.

Yes. Metrobloks is specifically designed for the power densities required by modern GPU clusters. AI training workloads running NVIDIA H100, H200, or AMD Instinct accelerators require 10–30 KW per server, and a full rack of GPUs can exceed 100 KW. Metrobloks liquid-to-chip cooling infrastructure supports 150 KW or more per rack, making it one of the few colocation providers genuinely capable of supporting large-scale AI infrastructure at full density.