Q.ANT Unveils Photonic NPU Gen 2+ in Chicago

The optical co-processor performs Discrete Fourier Transforms directly in light, debuting at the SC26 conference.

Updated on Nov. 17, 2026 in Quantum Computing

Close view of a small photonic chip with fine glowing light lines etched across its surface, resting on a dark metal plate with a thin cyan beam of light.
Q.ANT unveiled its NPU Gen 2+ at SC26 in Chicago, adding optical Discrete Fourier Transforms to speed AI computations directly in light. AI Illustration. Upload story photo >

Q.ANT unveiled its NPU Gen 2+ at SC26 in Chicago, adding optical Discrete Fourier Transforms to its Native Processing Unit technology. The company also demonstrated the chip paired with an AMD GPU to run a large language model.

Why it matters

Advanced functions like the Discrete Fourier Transform allow AI models to work with fewer parameters and less data, while digital hardware spends significant time and energy computing the transform at high data volumes. Performing the operation in light on a photonic chip targets that bottleneck directly.

The NPU Gen 2+ performs Discrete Fourier Transforms directly in light on a photonic chip and remains rack-mountable through PCIe integration into standard servers. The source provides no measured throughput or energy-efficiency numbers for the chip.

The players

Q.ANT

A photonic computing company founded in 2018 with its European headquarters in Stuttgart, Germany and offices in Munich and Austin, Texas.

AMD

A major semiconductor company whose GPU was paired with Q.ANT's chip in the large language model demonstration.

IONOS

A European cloud and hosting provider that partnered with Q.ANT on a photonic pilot giving access to real hardware.

The details

At SC26, Q.ANT showed the NPU Gen 2+ as a co-processor working alongside an AMD GPU to run a large language model using optical Fourier transforms. The company also launched an open-source SDK in September 2026 with C and Python APIs, example applications, documentation and a CPU-based simulation environment, plus a photonic pilot with IONOS giving access to real hardware.

Timeline

  1. 2025: Q.ANT launched the NPU Gen 2 and secured Europe's largest Series A for photonic computing.

  2. September 2026: Q.ANT launched its open-source software development kit.

  3. November 17, 2026: The NPU Gen 2+ was unveiled at SC26 in Chicago.

  4. Second half of 2027: Q.ANT plans to offer commercial systems combining Gen 2 and Gen 2+ processors.

The Tech Race

The Gen 2+ extends the NPU Gen 2 platform launched in 2025 by adding optical Discrete Fourier Transforms, positioning Q.ANT's photonic approach as an alternative to purely digital AI accelerators. Q.ANT's planned second-half 2027 commercial systems will combine both generations.

For developers, the open-source SDK with C and Python APIs and a CPU-based simulation environment allows experimentation before access to physical hardware. The IONOS photonic pilot offers a route to run workloads on real photonic hardware without owning it.

The takeaway

Photonic co-processors are moving from lab demonstrations toward commercial systems that pair with standard GPUs and servers. Readers following AI hardware should watch whether optical transforms deliver measurable efficiency gains once the planned 2027 systems ship.

What happens next

Q.ANT plans to offer commercial systems combining Gen 2 and Gen 2+ processors in the second half of 2027.

Further reading

For background on the wider field, explore our Quantum Computing coverage.