Partnerships
Q.ANT Opens Photonic AI Hardware Pilot on IONOS Bare Metal
Add Unite.AI to your preferred sources on Google
Q.ANT announced on November 9, 2026 that selected applicants can test their own AI workloads on its second-generation photonic processor through dedicated IONOS bare-metal servers in Germany, in a free closed-beta pilot intended to identify commercial uses.
The program is limited to participants with a concrete use case and enough expertise to administer a bare-metal server independently. IONOS manually provisions each system, with SSH root access and an optional KVM console. Its pilot documentation describes the ISO 27001-certified German hosting as compliant with the GDPR data-protection regime and sovereign within the EU.
From Simulation to Physical Photonic Hardware
The pilot extends a development path Q.ANT began with its open-source software development kit on September 23, 2026. That release supplied programming interfaces for C and Python, sample applications, documentation and a CPU-based simulation backend; the new program lets approved users move those applications onto physical Native Processing Units (NPUs). Q.ANT provides drivers, its software stack and PyTorch integration, with model and framework compatibility settled during onboarding.
The NPU executes nonlinear functions directly in the optical domain rather than decomposing them into long sequences of digital steps on conventional transistor-based chips. Q.ANT says this approach can yield comparable or better results with fewer parameters, operations and resources. Accepted workloads span industrial vision, time-series forecasting, scientific simulation and robotics, as well as large-language-model and foundation-model inference, physical AI and world models.
Q.ANT’s pilot page reports 40% fewer parameters for image processing and 65% fewer for time-series prediction at the same accuracy, results the closed beta is designed to test against users’ own data and software. That range of workloads gives Q.ANT a way to establish where optical nonlinear processing is useful beyond its initial research deployments.
A Structured Evaluation for Commercial Use
After an application review and video call, participants receive tutorials, code samples and documentation, followed by workload reviews and adjustments to the test focus. A final workshop examines energy and performance metrics, suitability for production and a possible next phase. Participants retain ownership of their workloads, data and results, which remain confidential; only aggregated, anonymized insights from discussions feed into processor and platform development.
The pilot follows the May 19, 2026 commercial agreement that made IONOS the first commercial customer for Q.ANT’s Native Processing Server. Q.ANT positions the NPU as a co-processor for particular AI workloads alongside CPUs and GPUs, rather than a wholesale GPU replacement. The closed beta therefore adds customer-run evaluation to an existing commercial relationship rather than introducing the partnership itself.
Independent Tests Show Gains and Constraints
An April 21, 2026 evaluation by the Leibniz Supercomputing Centre provides a separate reference point. LRZ tested both NPU generations on AI and high-performance computing (HPC) tasks and reported second-generation chips running up to 50 times faster in the best case, with a 25-fold acceleration in image recognition using a convolutional neural network. For typical HPC workloads, power consumption fell by roughly 50% to 84% from the first to second generation, depending on computational complexity.
The evaluation also identified limits relevant to the pilot. The co-processors cannot perform every operation optically and must exchange data continuously with classical processors. Energy efficiency is highest when computations remain on the optical unit, making data movement a central design constraint. The analog system converts data at 16-bit precision and must account for photodetector noise; more precise results require complex tasks to be split into multiple operations and more sophisticated software tools. LRZ researchers estimated that the chips could be ready for AI-cluster deployment after one or two further development cycles.
The pilot now puts those questions in customers’ hands: whether the reported gains survive real workloads, data movement and software constraints will determine where photonic co-processing fits beside CPUs and GPUs in production AI infrastructure.
