Latest top stories
Start-ups
Technology

Light Over Silicon: Q.ANT’s Big Bet on Sustainable AI Compute

8 December 2025

As AI infrastructure races ahead, the physics of conventional chips are quietly becoming a hard constraint. Data centres struggle with power limits, cooling overhead and the growing political scrutiny around their environmental footprint.

Into that pressure cooker steps Stuttgart-based Q.ANT, a photonic computing scale-up aiming to reshape sustainable AI infrastructure by replacing electrons with light in the data centre — and now has the funding, manufacturing footprint and first reference deployments to try.

In the past year, the company has closed Europe’s largest photonic computing Series A, launched a dedicated photonic chip production line, deployed systems at a German supercomputing centre, announced a multi-year partnership with Forschungszentrum Jülich, and unveiled its second-generation photonic processor, positioning itself as a key player in energy-efficient AI computing.

From TRUMPF Spin-off to Photonic Deep-Tech Scale-up

Founded in 2018 as a spin-off from German industrial group TRUMPF, Q.ANT develops photonic processing systems built on thin-film lithium niobate (TFLN), enabling computation “natively with light.” Its Light Empowered Native Arithmetics (LENA) architecture allows nonlinear mathematical operations to be executed optically, delivering:

  • dramatically reduced heat generation

  • higher computational density

  • and potentially major improvements in performance per watt

In late 2024, Q.ANT introduced its first commercial Native Processing Unit (NPU) on a PCIe accelerator card, marketed with 30× higher energy efficiency for select AI workloads. That NPU now forms the basis of the company’s Native Processing Server (NPS) — a 19-inch rack-mountable photonic co-processor designed to slot into existing data centre infrastructure.

For organisations exploring lower-carbon AI, the pitch is simple: faster AI, dramatically lower energy use, no new data-centre architecture required.

Europe’s Largest Photonic Computing Funding: €62M + Duquesne Extension

In July 2025, Q.ANT announced a €62 million Series A — Europe’s largest fundraising round in photonic computing. Investors included Cherry Ventures, UVC Partners, and imec.xpand, supported by L-Bank, Verve Ventures, Grazia Equity, EXF Alpha, LEA Partners, Onsight Ventures and TRUMPF.

The capital is aimed at:

  • scaling the manufacturing of Q.ANT’s photonic chips

  • accelerating commercial roll-out of the NPS servers

  • extending the technical roadmap for photonic AI hardware

  • and enabling international expansion, particularly in the U.S.

In October 2025, the Duquesne Family Office joined the round, bringing total Series A financing to US$80 million. That extension further solidified Q.ANT’s position as a leading contender in the global race for next-generation sustainable AI hardware.

Upcycling a CMOS Line: A New Model for European Chip Sovereignty

Manufacturing has historically been the Achilles heel of photonic computing. Q.ANT’s response: repurpose existing semiconductor infrastructure instead of building new fabs from scratch.

In early 2025, Q.ANT and IMS CHIPS launched a dedicated production line for TFLN-based photonic AI chips by upcycling an established CMOS facility. The result:

  • capacity for ~1,000 wafers per year

  • a shorter path from R&D to commercial chips

  • and a blueprint for strengthening European chip sovereignty

This approach supports two critical objectives at once: sustainable AI compute and a more resilient, regionally anchored semiconductor supply chain.

From Lab Demo to Supercomputer: Deployment at LRZ

In July 2025, Q.ANT’s photonic Native Processing Server was deployed at the Leibniz Supercomputing Centre (LRZ) — marking the first integration of a photonic co-processor into an operational European HPC environment.

LRZ is testing the hardware on:

  • AI inference

  • computer vision

  • scientific workloads, including physics simulations

Q.ANT highlights potential 90× lower power consumption for specific tasks, offering an early glimpse at how photonic computing could reshape energy-efficient HPC. German federal funding frames this as part of a national strategy for future-proof, sustainable supercomputing.

Jülich Partnership: A Four-Year Photonic HPC Testbed

In November 2025, Q.ANT announced a multi-year collaboration with the Jülich Supercomputing Centre (JSC). The partnership will evaluate how photonic processors can accelerate key HPC kernels and AI routines within JSC’s Modular Supercomputing Architecture.

This collaboration serves two purposes:

  1. Technical validation of photonic processors in demanding scientific environments

  2. Algorithmic co-design, ensuring new hardware aligns with emerging AI/HPC workflows

With both LRZ and Jülich on board, Q.ANT now has two of Europe’s premier HPC institutions acting as early evaluators — a powerful signal of credibility.

NPU 2: Second-Generation Photonic AI Processor

In November 2025, Q.ANT launched NPU 2, its next-generation photonic processor. The company reports:

  • enhanced nonlinear optical processing

  • significant gains in performance per watt

  • and a turnkey rack-mounted server for easy deployment

The NPU 2 server integrates multiple photonic NPUs with an x86 host, Linux OS and standard programming interfaces in C/C++ and Python — making it easier for AI teams to experiment without rewriting their entire software stack.

Customer shipments are planned for early 2026, indicating that Q.ANT is entering a genuine commercialisation phase.

Performance Claims, Real-World Questions

Q.ANT routinely cites potential gains of:

  • 30× energy efficiency

  • 50× performance improvements

  • up to 100× higher compute density

These figures are tied to specific workloads and internal benchmarks. Large-scale independent validation is still to come, and will ultimately determine how transformative photonic computing can be in everyday AI infrastructure.

Still, the direction of travel is unmistakable: radically more energy-efficient AI is technically achievable, and photonic approaches are emerging as one of the strongest candidates.

Why Q.ANT Matters for Sustainable AI — and for Europe

From the perspective of sustainable AI and strategic tech independence, Q.ANT sits at the centre of three converging trends:

1. AI’s energy crisis

As AI workloads explode, energy efficiency has shifted from “nice to have” to “non-negotiable.” Photonic processors promise a meaningful reduction in carbon footprint per model, especially in inference-heavy environments.

2. Upcycled manufacturing

Q.ANT’s fab strategy demonstrates how legacy semiconductor facilities can be modernised for new compute paradigms — a crucial capability for Europe.

3. A European foothold in next-gen compute

If photonic computing becomes a defining architecture for sustainable AI, Europe has a rare opportunity to lead a strategic hardware category rather than chase one.

Conclusion: Computing With Light Has Left the Lab

Q.ANT is no longer a speculative photonics start-up. With $80M in funding, a functioning TFLN production line, deployments in major HPC centres, and the launch of its second-generation processor, the company has become one of the most credible contenders for energy-efficient AI computing.

As AI power demand continues to surge, Q.ANT’s proposition — faster computation, drastically lower energy use, and compatibility with existing data centres — positions it squarely within the most urgent conversation in the industry: how to scale AI sustainably.

For now, one thing is clear:
computing with light is no longer a futuristic promise — it’s becoming a commercial reality.