# Photonic Native Processing Unit by Q.ANT Canonical page: https://technooptimist.io/computing/native-processing-unit Frontier: [Computing](https://technooptimist.io/computing.md) · Field: New kinds of computers · Company: [Q.ANT](https://technooptimist.io/companies/q-ant.md) Stage: Pilot (3 of 5: Research → Proto → Pilot → Deploy → Scaling) Updated: last update 2026-05-19, facts checked 2026-10-11 Goal: Cut the energy and heat of AI computing by doing its heaviest maths with light, which carries out multiplications without the switching losses of billions of transistors. Status: Second generation running at LRZ and Jülich; IONOS rolling it out to cloud customers later in 2026. Next milestone: IONOS cloud rollout to business users, target 2026 ## Milestones | Status | Milestone | Date | Slip | | --- | --- | --- | --- | | Done | Founded as a TRUMPF spin-off | 2018 | | | Done | Photonic chip pilot line opens at IMS CHIPS | Feb 2025 | | | Done | €62M Series A | 18 Jul 2025 | | | Done | Jülich partnership; JSC buys a server | 18 Nov 2025 | | | Done | Second-gen NPUs deployed at LRZ | 18 Mar 2026 | | | Done | IONOS signs as first commercial customer | 19 May 2026 | | | Current | IONOS cloud rollout to business users | 2026 | | ## Upcoming events - **2026**: IONOS cloud rollout to business users ## Most important updates - **19 May 2026**: IONOS becomes Q.ANT's first commercial customer for photonic AI servers. European cloud provider IONOS will deploy Q.ANT's Native Processing Server later in 2026 and offer it to its business users, the first commercial sale after deployments at LRZ and Jülich. ([our story](https://technooptimist.io/stories/ionos-becomes-q-ant-s-first-commercial-customer-for-photonic-ai-servers-797aeb39.md) · [original source](https://qant.com/news/q-ant-takes-photonic-ai-computing-commercial-as-ais-power-demand-surges/)) - **18 Mar 2026**: Q.ANT deploys second-generation photonic processors at the Leibniz Supercomputing Centre. Q.ANT installed second-generation NPUs at LRZ under a German research ministry project on hybrid digital-analog HPC. ([our story](https://technooptimist.io/stories/q-ant-deploys-second-generation-photonic-processors-at-the-leibniz-5b6b2662.md) · [original source](https://bebeez.eu/2026/03/18/photonic-chip-company-q-ant-deploys-second-gen-photonic-processors-at-germanys-lrz/)) - **18 Nov 2025**: Jülich Supercomputing Centre and Q.ANT start a four-year photonic computing partnership. JSC will buy a Q.ANT Photonic Native Processing Server, integrate it into its modular supercomputing setup and develop algorithms suited to computing with light. ([our story](https://technooptimist.io/stories/julich-supercomputing-centre-and-q-ant-start-a-four-year-photonic-467be46c.md) · [original source](https://www.photonicsonline.com/doc/computing-light-forschungszentrum-julich-q-ant-launch-collaboration-photonic-computing-0001)) - **18 Jul 2025**: Q.ANT raises €62M Series A to scale photonic AI processors and expand to the US. Q.ANT raised €62M ($72M), co-led by Cherry Ventures, UVC Partners and imec.xpand, to scale production of its photonic processors and build next-generation chips. ([our story](https://technooptimist.io/stories/q-ant-raises-62m-series-a-to-scale-photonic-ai-processors-and-expand-to-94e1c3d8.md) · [original source](https://www.verdict.co.uk/q-ant-funding-photonic-ai-processors/)) ## Current obstacles - **From pilot to volume**: Chips come from a pilot line; supplying cloud customers like IONOS needs higher-volume, high-yield production. - **Workload fit**: Gains apply to operations that map onto light well; whole AI applications still run mostly on GPUs. ## Physics limits - **Analog precision is limited**: Light intensity carries noise from lasers and detectors, so analog optical maths gives a few bits of precision, enough for many AI steps but not for exact arithmetic. - **Converting costs energy**: Every number must be turned into light and back. Those converters can eat much of the saving unless enough maths is done in light per conversion. - **Light is big**: Optical components are limited by the wavelength of light, around a micron, so photonic circuits pack far fewer operations per square millimetre than transistors. ## How it works ### Encode: Numbers become light Input values are written onto laser light by modulators that change its brightness or phase in proportion to each number. ### Compute: Waveguides do the maths Light passes through a network of on-chip waveguides and modulators that multiply and add the signals as they travel, with almost no heat produced on chip. ### Read out: Back to digital Photodetectors turn the result back into electrical signals, which are digitised and returned to the host over PCIe. ### Role: A co-processor, not a GPU The server takes specific heavy operations from CPUs and GPUs, as in climate models, medical imaging and fusion simulations at LRZ. ## Spec sheet | Spec | Photonic Native Processing Unit | Provenance | | --- | --- | --- | | Gen 2 vs gen 1 (LRZ evaluation) | Up to 100x performance | reported | | Matrix-multiply throughput (LRZ) | Over 50x higher, 6x less energy | reported | | Energy efficiency claim | Up to 30x vs conventional | reported | | Host interface | Standard PCIe | reported | | Chip platform | Thin-film lithium niobate (TFLN) | reported | Provenance: reported = stated by the company; estimated = our estimate; sample = a sample figure. ## About Q.ANT Q.ANT is a Stuttgart company, spun out of the machine maker TRUMPF, that builds processors doing AI maths with light instead of electrons. Its servers plug into ordinary data-centre racks as co-processors. - Second-gen photonic NPU running at the LRZ and Jülich supercomputing centres; IONOS first commercial buyer (May 2026). - Raised €62M Series A in July 2025, Europe's largest for photonic computing. - Opened a US headquarters in Austin in April 2026. - Founded: 2018 - Headquarters: DE - Status: private - Raised: $72M - Last round: Series A $72M, 18 Jul 2025, led by Cherry Ventures, UVC Partners, imec.xpand - Website: https://qant.com - People: Michael Förtsch (Founder & CEO); Bruno Spruth (CTO) - Partners, customers, investors: IONOS (customer); Leibniz Supercomputing Centre (LRZ) (customer); Forschungszentrum Jülich (partner); TRUMPF (investor) Company page: https://technooptimist.io/companies/q-ant.md ## Related programs Everything in Computing: https://technooptimist.io/computing.md --- The Techno Optimist tracks every frontier of technology, program by program: https://technooptimist.io/ (a Markdown version of any page: add .md to its URL; site map for agents: https://technooptimist.io/llms.txt). Data API (JSON, realtime stream): https://technooptimist.io/developers. People read 5 pages free, then Reader $10/month; these Markdown pages are open to agents.