> ## Content Index
> Fetch the complete content index at: https://nexi.fund/llms.txt
> Use this file to discover other available public pages before exploring further.

# Photon Bridge's €10M Bet: The Laser Is AI's Next Supply-Chain Bottleneck
- URL: https://nexi.fund/photon-bridge-laser-bottleneck-2026/
- Published: 2026-10-23T07:00:27.000Z
- Updated: 2026-10-23T07:00:27.000Z
- Description: A 15-person Dutch team raised €10 million to industrialise the light source inside co-packaged optics. The argument is about manufacturing yield, not optical power — and the 2028 volume claim is still two years from being testable.
- Author: Nexi.fund Labs
- Tags: AI & Infrastructure, #mode-2, #hook-character, #track-A, #layout-sparse-brief

Paul Marchal took over as CEO of Photon Bridge in September 2025\. The pitch he has been making since is a manufacturing argument, not an optics one. The Eindhoven company does not build transceivers or switches. It builds the light source that feeds them: laser arrays multiplexed onto a silicon interposer and packaged as an external laser source that sits beside a co-packaged optics engine.

On 21 October the company said it raised €10 million in a seed round co-led by Amadeus APEX Technology Fund and 360 Capital. Innovation Industries, Faber and Brabant Development Agency increased their existing investment. Photon Bridge employs 15 people and expects to reach 20 by the end of 2026.

The cheque is small for a hardware company. The claim attached to it is not: the laser is becoming the part of AI networking that cannot be made fast enough.

## Why the laser count outruns the network

Optical links carry data between accelerators because copper runs out of reach and power at high bandwidth. Co-packaged optics (CPO) shortens the electrical path further by moving the optical engine next to the switch or compute die. Each engine still needs continuous light. That light comes from a separate laser module, and every wavelength in the link needs its own emitter.

Add channels, and the laser count climbs with the channel count instead of the port count. Omdia expects more than 20 million data-centre AI accelerators to ship annually by 2028, according to the funding announcement. Goldman Sachs Research expects the optical networking market to grow ninefold to $154 billion over the same period.

Photon Bridge argues that supply is the mismatch. High-power lasers for telecom and subsea cables come off lines built for thousands of units, and the company says AI now demands millions. On its account, that gap between telecom volumes and AI volumes sets the pace of deployment, with optical performance a secondary constraint.

## Many modest lasers on one interposer

Most external laser sources chase power per emitter. Photon Bridge inverts that. It runs distributed-feedback (DFB) laser arrays at lower power density and combines 8, 16 or 32 wavelengths onto a single fibre through multiplexing built into the silicon interposer.

Lower power density makes each laser less sensitive to fabrication defects. Yield improves, and yield matters more as volume rises. The design target is up to 40 milliwatts per wavelength per fibre, produced across several indium phosphide (InP) foundries.

Assembly carries the other half of the argument. A patented cantilever waveguide interface absorbs the roughly 2-micron placement tolerance of standard flip-chip bonding and turns it into sub-200-nanometre optical alignment, with no active alignment step. Thick-film silicon waveguides handle more than 1 watt on chip without the nonlinear losses that thinner platforms hit.

In February 2026 the company demonstrated more than 30 milliwatts per channel at the chip edge in continuous-wave operation at room temperature, which it frames against the per-channel requirements of 1.6T and 3.2T CPO engines. In March it named the Canadian Photonics Fabrication Centre as its InP foundry partner. In April it added PHIX for packaging.

> Instead of making individual lasers increasingly powerful and difficult to manufacture, we move that scaling challenge into the architecture.— Paul Marchal, CEO, Photon Bridge

## What €10 million has to buy

The round funds customer qualification. First samples are due in 2027, with volume manufacturing prepared from 2028\. That timeline is the honest part of the pitch and the main risk. A sample in 2027 leaves roughly a year to qualify a component inside someone else's CPO module before the 2028 volume claim gets tested.

Photon Bridge is fabless. It designs the platform and leans on commercial foundries and outsourced assembly and test partners. Capital needs stay low. Control over the yield curve it promises to fix does not.

📊

**Three things the next twelve months must show**  
  
A named CPO or module customer, not a partnership announcement  
  
Per-channel power holding near 40 mW across all three channel configurations  
  
A second InP foundry qualified for volume, not just samples 

Competition is not idle. Ayar Labs, Lumilens and Celestial AI are raising against the same interconnect bottleneck, and hyperscalers are designing more of their own optics. Photon Bridge's differentiation is narrow by design: the light source, not the link.

The narrowness cuts both ways, and the market will price it soon enough. A component supplier with no public design wins is a cheap option on a large market. €10 million buys qualification time. It does not buy capacity, and it does not yet buy a customer.

As we wrote in [October](https://nexi.fund/volantis-photonic-memory-series-a-2026/), photonics capital has been moving toward the layers beneath the transceiver. Photon Bridge is asking investors to fund the layer beneath that.

[ Photon Bridge raises €10M to attack the AI laser bottleneck Company release with the round terms, the Omdia and Goldman Sachs market figures, and the 2027 sample / 2028 volume roadmap. Photon Bridge ](https://www.photonbridge.com/news/photon-bridge-raises-%E2%82%AC10m-to-tackle-ai%E2%80%99s-emerging-laser-bottleneck?ref=nexi.fund) 

Primary source. Treat the market forecasts as company-selected third-party projections.

[ Dutch photonics startup Photon Bridge raises €10M Trade coverage carrying the full investor quotes from Marchal, Hauer and Maifredi, plus the Omdia accelerator forecast. Tech.eu ](https://tech.eu/2026/10/21/dutch-photonics-startup-photon-bridge-raises-eur10m-to-tackle-ais-looming-laser-bottleneck/?ref=nexi.fund) 

Independent confirmation of round size and co-leads.

[ Photon Bridge raises €10m for chip-to-chip laser devices Sifted's exclusive, which labels the round a seed and frames the product as laser hardware for connecting AI chips in data centres. Sifted ](https://sifted.eu/articles/photon-bridge-laser-device-ai-chips-seed-round?ref=nexi.fund) 

Round-stage confirmation: seed, not a priced Series A.