Summary
- Selected participants can test Q.ANT's second-generation photonic processors through IONOS bare-metal infrastructure.
- The pilot supports existing software tools and evaluates performance using real customer workloads.
- Efficiency benefits remain subject to workload-specific testing rather than universally demonstrated performance.
Q.ANT and IONOS have launched a commercial pilot allowing selected developers to test artificial intelligence and scientific computing workloads on photonic processors installed within a German data centre.
The programme, introduced on 9 November, provides access to Q.ANT’s second-generation Native Processing Unit (NPU) through dedicated bare-metal servers operated by IONOS.
Participants will be able to use their own data and software to evaluate how effectively the optical processors execute computational tasks that are normally performed using conventional electronic hardware.
The collaboration follows earlier evaluations of Q.ANT processors in German high-performance computing environments and extends hardware access to selected commercial users.
Q.ANT’s technology uses photonic integrated circuits to perform particular mathematical operations using light. Its architecture is designed to handle nonlinear functions and other operations relevant to artificial intelligence, scientific simulation, and industrial computing.
The potential advantage is a reduction in the number of conventional digital operations required for suitable workloads. Whether that translates into lower total energy consumption depends on the complete computing system rather than the optical processor alone.
The programme will be available without charge to a limited number of accepted applicants. Participants must submit a defined use case and demonstrate sufficient technical expertise to operate the dedicated server environment.
Hardware access will be provided through an ISO 27001-certified IONOS data centre in Germany. The specific facility has not been publicly identified.
Q.ANT will supply the processor, software stack, and technical support. IONOS will provide the hosting infrastructure and associated server operation.
Performance testing moves beyond simulation
The pilot is designed to establish how photonic processing performs when used for practical customer applications rather than simulated workloads or isolated processor benchmarks.
Developers can work with tools including C/C++, Python, and PyTorch integration through Q.ANT’s software development kit.
The technology is particularly suited to testing mathematical functions that can be implemented directly in the optical domain, including selected nonlinear transformations and matrix operations.
Potential applications include image processing, robotics, time-series prediction, scientific modelling, and AI inference.
Each selected project will begin with technical onboarding and a review of the proposed workload. Participants will then work with Q.ANT to examine performance before completing a final assessment of the results.
The programme is also intended to identify workloads for which the processor provides little advantage, an important consideration when evaluating specialised computing architectures.
Hardware that performs particular mathematical functions efficiently may still provide limited overall improvement if an application remains constrained by data transfer, memory access, or conventional processing stages.
Any meaningful comparison must therefore consider completed workload throughput, energy consumption, accuracy, and the resources required to operate the surrounding server system.
Q.ANT has previously published performance and efficiency claims based on selected demonstrations. Those figures cannot be treated as representative of every application or as independent confirmation of production-scale savings.
The current programme creates an opportunity to obtain further measurements using workloads selected by participating organisations.
Data centre implications depend on system performance
The development is relevant to computing infrastructure because rising processor power requirements place increasing demands on electricity distribution and cooling systems.
Accelerators that perform useful calculations using less energy could potentially reduce some of those requirements, although the practical effect depends on how the equipment is integrated into servers and racks.
Photonic processors still require electrical interfaces, memory access, data movement, and supporting control hardware.
Even where optical computation itself is highly efficient, surrounding electronic components continue to consume power and generate heat.
Consequently, a processor-level improvement does not automatically establish a proportional reduction in the electricity or cooling capacity required by a data centre.
Q.ANT has already supplied processors for evaluation at the Leibniz Supercomputing Centre and established a presence at the Jülich Supercomputing Centre. Those earlier projects provide operational experience, while the IONOS programme introduces a commercial hosting environment.
The pilot begins during November, with participant feedback expected to inform further processor and software development.
Neither partner has disclosed the number of servers allocated to the programme or a date for general commercial availability.
Further deployment will depend on the results achieved with actual applications and whether the processor can deliver sufficiently repeatable improvements to justify its use within conventional computing infrastructure.
