Summary
- The proposed single-tenant cluster comprises 128 Nvidia HGX B300 servers and 1,024 B300 Blackwell Ultra GPUs.
- Vertical Data puts the contemplated five-year contract value at approximately $192.1 million, with Q1 2027 service targeted.
- Take-or-pay and deposit obligations are described as binding, while most remaining terms still require a definitive Master Services Agreement.
Vertical Data has signed a letter of intent for a 1,024-GPU Nvidia Blackwell Ultra cluster at an unnamed Western European data centre, with the proposed five-year agreement valued by the company at approximately $192.1 million.
The proposed deployment would use 128 Nvidia HGX B300 servers containing 1,024 B300 Blackwell Ultra GPUs and would be dedicated to a single unnamed global AI developer. Vertical Data is targeting service commencement in the first quarter of 2027.
The announcement is not a completed five-year infrastructure contract. Vertical Data says the LOI contains binding take-or-pay and deposit obligations covering the contracted compute and storage capacity, but most of the remaining commercial, operational, and delivery terms are non-binding and still subject to negotiation of a definitive Master Services Agreement.
That distinction is central to the scale being presented. The $192.1 million figure is the contemplated aggregate value across the initial five-year term. Completion remains dependent on the final agreement as well as financing, hardware allocation, regulatory compliance, due diligence, construction, and delivery of the associated data centre infrastructure.
If completed as described, the cluster would be engineered for sovereign AI workloads within the European Economic Area. Vertical Data says the deployment will include physical and network isolation, multi-petabyte storage, 24-hour managed engineering support, and controls intended to support GDPR, ISO 27001, and SOC 2 Type II requirements.
The location has not been disclosed beyond Western Europe. That leaves an important part of the infrastructure picture unresolved, because a 1,024-GPU Blackwell cluster requires more than access to accelerators. The delivery programme has to align server hardware, electrical capacity, cooling, network connectivity, storage, security, and the data centre’s own construction or fit-out timetable.
Vertical Data operates across GPU provisioning, financing, and data centre interests. The company says its Vertical Edge business holds equity in facilities it sources, develops, leases, and manages, while separate units provide GPU infrastructure and structured financing.
That integrated model reflects the way large AI deployments increasingly cut across what were previously separate procurement decisions. A customer seeking high-density capacity may need to secure scarce accelerators and the facility capable of powering and cooling them on the same programme.
Blackwell-generation systems also increase the importance of facility readiness. AI clusters concentrate substantial electrical load into relatively small footprints, and operators have been moving towards higher-density rack designs and liquid cooling to manage processors that cannot be accommodated efficiently through conventional air cooling alone.
The proposed Q1 2027 start therefore leaves a relatively short implementation window. Vertical Data says it intends to use a priority hardware pipeline, but the company itself identifies infrastructure construction, GPU delivery, financing approval, supplier allocation, export controls, sanctions compliance, and due diligence among the conditions that could affect completion.
It also has not named the customer. An unnamed counterparty makes independent assessment of the underlying demand and credit profile difficult, while the fact that a substantial majority of the transaction remains subject to the MSA means the project should be treated as a proposed deployment rather than contracted operating capacity.
The binding provisions described by Vertical Data do give the LOI more substance than a purely indicative memorandum. Take-or-pay structures are designed to commit a customer to payment for reserved capacity whether or not it fully consumes that capacity, giving an infrastructure provider greater certainty when financing expensive hardware and facilities.
Whether that structure ultimately supports the full $192.1 million deployment will depend on the definitive agreement and physical delivery. The project joins a growing group of European AI infrastructure proposals in which compute procurement, sovereign data requirements, and data centre capacity are being contracted together rather than sequentially.
The next material milestone is therefore not another headline valuation but execution of the MSA. Until that happens, the 1,024-GPU cluster represents a substantial proposed Western European deployment with some binding commercial provisions — not a fully completed infrastructure contract.
