Q.ANT Opens Free Photonic AI Chip Pilot on IONOS Servers in Germany

Q.ANT will let selected applicants run their own AI workloads on its second-generation photonic processor through dedicated IONOS bare-metal servers in Germany, the company announced on November 9, 2026. The closed-beta pilot is free of charge and explicitly aimed at identifying commercial use cases for optical AI hardware.

The stakes are straightforward. Photonic processors promise large energy and speed gains over transistor-based chips, but the field has few customer-run deployments to prove those claims on real data. This pilot tests whether laboratory numbers survive contact with production workloads.

What does the pilot actually offer?

Participants get access to physical Native Processing Units (NPUs), Q.ANT's term for its photonic co-processors, hosted on IONOS bare-metal servers (dedicated physical machines, not virtualized, provisioned manually by IONOS). Each system ships with SSH root access and an optional KVM console for out-of-band management.

The program is deliberately selective. Q.ANT restricts participation to applicants with a concrete use case and sufficient expertise to administer a bare-metal server independently. The pilot documentation describes the hosting as ISO 27001-certified, compliant with the GDPR data-protection regime, and sovereign within the EU — a positioning point for European customers weighing where their AI compute runs.

How did the hardware get here?

The pilot extends a development path that began on September 23, 2026, when Q.ANT released an open-source software development kit. That release included:

• Programming interfaces for C and Python
• Sample applications and documentation
• A CPU-based simulation backend

The new program moves approved applications from simulation onto physical NPU hardware. Q.ANT supplies drivers, its full software stack and PyTorch integration, with model and framework compatibility settled during onboarding rather than guaranteed in advance.

The NPU executes nonlinear functions directly in the optical domain instead of decomposing them into long sequences of digital steps on conventional chips. Q.ANT claims this approach can deliver comparable or better results with fewer parameters, operations and resources. Accepted workload categories span:

• Industrial vision
• Time-series forecasting
• Scientific simulation
• Robotics
• Large-language-model and foundation-model inference
• Physical AI and world models

What efficiency numbers does Q.ANT report?

The numbers lead the pitch. Q.ANT's pilot page reports 40% fewer parameters for image processing and 65% fewer for time-series prediction at the same accuracy as conventional implementations. The closed beta exists precisely to test those figures against users' own data and software, not Q.ANT's benchmarks.

That workload breadth matters strategically: it gives Q.ANT a structured way to establish where optical nonlinear processing is useful beyond its initial research deployments.

How does the evaluation process work?

The pilot follows a defined sequence. After an application review and a video call, participants receive tutorials, code samples and documentation. Q.ANT then conducts workload reviews and adjusts the test focus. A final workshop examines three questions:

• Energy and performance metrics
• Suitability for production use
• A possible next phase of the partnership

Intellectual-property terms favor the participant. Users retain ownership of their workloads, data and results, all of which remain confidential. Only aggregated, anonymized insights from discussions feed back into Q.ANT's processor and platform development.

The commercial backdrop dates to May 19, 2026, when IONOS signed on as 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, not a wholesale GPU replacement. The closed beta therefore adds customer-run evaluation to an existing commercial relationship rather than announcing a new partnership.

What do independent tests show?

A separate reference point comes from the Leibniz Supercomputing Centre (LRZ), which published an evaluation on April 21, 2026. LRZ tested both NPU generations on AI and high-performance computing tasks and reported:

• Second-generation chips running up to 50 times faster in the best case
• A 25-fold acceleration in image recognition using a convolutional neural network
• Power consumption for typical HPC workloads falling roughly 50% to 84% from the first to second generation, depending on computational complexity

Where are the limits?

The LRZ evaluation also identified constraints directly relevant to pilot participants. The co-processors cannot perform every operation optically and must exchange data continuously with classical processors. Energy efficiency peaks when computations stay on the optical unit, making data movement the central design constraint.

The analog system introduces further precision trade-offs. It converts data at 16-bit precision and must account for photodetector noise. Achieving more precise results requires splitting complex tasks into multiple operations and building more sophisticated software tools. LRZ researchers estimated the chips could be ready for AI-cluster deployment after one or two further development cycles.

What comes next?

The pilot now places those open questions in customers' hands. Whether Q.ANT's reported gains survive real workloads, data-movement bottlenecks and software constraints will determine where photonic co-processing fits beside CPUs and GPUs in production AI infrastructure. The closed beta, with its final workshop on energy, performance and production readiness, is the mechanism that will produce those answers.