Summary
- T-Systems and Dresden-based Sparxity have agreed to explore AI computing infrastructure integration.
- Initial work will use Sparxity's existing hardware to prepare for the next generation of specialised processors.
- A first deployment is targeted for 2028, but a definitive commercial installation and operating performance remain unconfirmed.
T-Systems and Dresden-based computing developer Sparxity have signed a letter of intent to integrate specialised AI hardware into European data centre infrastructure, targeting an initial deployment in 2028.
The agreement was signed at the German government’s Digital Summit and will bring Sparxity’s sparse computing technology together with T-Systems’ infrastructure and operational capabilities. The partners intend to evaluate how the hardware can be integrated into existing data centre environments before developing systems using a subsequent processor generation.
Sparxity, formerly SpiNNcloud Systems, is developing computing architecture designed to avoid unnecessary operations where workloads can be processed using sparse computation. T-Systems brings data centre infrastructure, system integration, and a wider partner network to the proposed collaboration.
The companies have not disclosed a definitive procurement contract, installation site, processor quantity, or committed electrical capacity. Their first deployment remains a target rather than an operating milestone, with detailed cooperation arrangements to be agreed over the coming months.
The engineering proposition centres on changing how computing resources are used. Some workloads include data or mathematical operations that can be omitted without changing the required result. Hardware designed to exploit those characteristics may reduce the amount of processing performed, but the benefit depends on the workload and its implementation.
Sparse computing does not guarantee the same efficiency improvement across all AI applications. Data movement, memory access, software design, and communication between processors can affect the performance and energy consumption of the complete system.
Testing the complete computing system
Introducing a new processor architecture into an operational data centre requires more than verifying the performance of an individual chip. Equipment must work with the supporting servers, network connections, management systems, electrical distribution, and cooling infrastructure.
Power consumption at processor level influences the heat that must be removed from the computing equipment. However, the resulting facility demand also depends on how densely hardware is installed, how efficiently it is used, and the energy consumed by supporting systems.
Where a computing architecture reduces the energy required for a specific workload, the facility may benefit from lower electrical and thermal demand for that work. Alternatively, operators could use the available power envelope to increase computing output, depending on their commercial and operating priorities.
The practical result cannot be determined without measurements under representative workloads. Neither partner has published verified site-level electricity savings, throughput figures, or efficiency comparisons arising from the proposed T-Systems integration.
The first phase will use existing Sparxity hardware to develop an approach for integrating its next generation of processors. This provides a route to examining software and hardware requirements before a more extensive installation is considered.
Data centre operational requirements also extend to support and maintenance. Specialist computing equipment needs procedures for fault detection, replacement, configuration, and compatibility with the services used to manage the installation.
Integration planning must address the availability and reliability of the complete computing environment, although the companies have not identified any particular unresolved fault or incompatibility in the systems under consideration.
A European infrastructure partnership
European governments and infrastructure providers are seeking greater control over computing capacity and associated technologies. The T-Systems and Sparxity cooperation brings together a European infrastructure operator and a German computing developer, but does not establish that every component of the eventual system will be manufactured or controlled within Europe.
Supply chains for advanced computing include semiconductor manufacturing, memory, networking, power electronics, software, and operational support. Their origins and contractual arrangements remain relevant when assessing claims about sovereignty and independence.
The agreement follows a separate Sparxity installation at Germany’s Jülich Supercomputing Centre, where specialised hardware has already been deployed for research workloads. That project provides technical context, but the new T-Systems programme is a distinct partnership with a later deployment objective.
T-Systems chief executive Ferri Abolhassan has identified the collaboration as part of the company’s effort to expand its AI infrastructure partner ecosystem. Hector A. Gonzalez, co-founder and chief executive of Sparxity, is participating in the agreement on behalf of the computing developer.
The planned 2028 deployment leaves a substantial validation and development period. The companies must still define the technical configuration, agree commercial arrangements, and establish how the proposed hardware performs within the infrastructure intended to support it.
Until those milestones are completed, the letter of intent represents an agreed direction for engineering cooperation rather than an addition to commissioned European computing capacity.
