Caprelion Opens 2027 Anchor AI Deployment Discussions for Planned ~2.5 MW High-Density Node in Slovakia

KOSICE, SLOVAKIA, September 1, 2026 /EINPresswire.com/ -- The Košice project is being developed as dedicated high-density AI infrastructure for operators planning new European GPU deployments, including NVIDIA GB300 NVL72-class systems. An early anchor operator may have the opportunity to align key power, liquid-cooling, network and operational requirements before the final technical configuration is frozen.

Europe’s race for AI infrastructure is increasingly about more than access to GPUs.

For operators planning the next generation of AI clusters, the critical question is becoming:
Where will power and high-density infrastructure actually be available when the next generation of GPU systems needs to be deployed?
Data-centre availability across Europe remains constrained. CBRE expects vacancy across primary and secondary European markets to fall to around 6% by the end of 2026, while new colocation signings associated with AI workloads more than quadrupled during the first half of the year.
JLL, meanwhile, has identified pre-leasing as an increasingly important part of occupier strategy, as contiguous blocks of high-density capacity become harder to secure in Europe’s established data-centre markets.

Against this backdrop, Caprelion is advancing a planned approximately 2.5 MW high-density AI infrastructure deployment in Košice, Slovakia, targeting a readiness window in the second half of 2027.
The company is beginning early discussions with potential anchor GPU infrastructure operators considering new European deployments for 2027–2028.
Infrastructure for an anchor deployment — not conventional rack-by-rack colocation
The Košice AI POD is not being positioned as a conventional data centre selling individual racks to a large number of unrelated customers.
The initial deployment is being developed around an anchor-scale high-density AI requirement. That changes the model materially.
Rather than asking the customer to fit into a completely predefined data-centre environment, Caprelion intends to consider the selected anchor operator’s requirements when defining key interfaces of the final infrastructure configuration.

These may include: GPU platform, rack density, electrical interfaces, direct liquid cooling, network fabric, storage architecture, security requirements, commissioning, acceptance and operating model. The value of early participation is therefore not limited to potential access to approximately 2.5 MW of future infrastructure.

It is also the opportunity to: influence the infrastructure envelope before design freeze.
An operator entering later in the process may instead need to adapt its compute platform to an increasingly fixed facility architecture. High-Density GPU Platforms, Including NVIDIA GB300. One architecture that illustrates how rapidly the requirements of modern AI data centres are changing is NVIDIA GB300 NVL72. NVIDIA describes GB300 NVL72 as a fully liquid-cooled rack-scale platform integrating 72 Blackwell Ultra GPUs and 36 Grace CPUs within a single system. For an infrastructure operator, this means that a modern AI deployment can no longer be treated simply as a collection of servers installed into conventional racks. The GPU architecture directly affects requirements for: rack power density, liquid cooling, hydraulic interfaces, electrical distribution, network fabric, physical layout and operations.
Caprelion’s technical qualification process is therefore intended to accommodate operators considering GB300-class deployments, as well as other high-density GPU architectures.
The current GPU-client qualification framework specifically requires the target platform class — GB200, GB300 or another architecture — to be identified before the project’s final technical configuration is established.
Caprelion does not represent the Košice project as already certified, approved or finally engineered by NVIDIA for GB300. Compatibility between a specific GPU configuration and the Košice AI POD infrastructure would need to be established through detailed engineering with the selected anchor operator and the relevant technology partners.

For an operator considering a European GB300 deployment in 2027, the advantage of entering the process early is precisely the opportunity to align these interfaces before final design freeze.

Caprelion
info@caprelion.com
Nodes Cluster, s.r.o.
Visit us on social media:
LinkedIn
YouTube