Photon Bridge has raised €10M to build laser light sources for optical links inside AI data centres, with first customer samples scheduled for 2027. The round was co-led by the Amadeus APEX Technology Fund and 360 Capital, the Eindhoven company said on Wednesday. Volume production is planned from 2028. The funding matters because optical connections have become a bottleneck for moving data between chips as AI clusters scale.
Funding round and 2027 sampling plan
Existing investors Innovation Industries, Faber and the Brabant Development Agency joined the round alongside the two lead funds. The company previously operated as Photon IP and raised €4.75M in seed funding last year before changing its name. The fresh capital will fund product development, customer testing and hiring in Eindhoven. Management describes a fabless model, working with outside chipmakers rather than running its own factory. It says lasers can be sourced from several suppliers, which reduces dependence on a single production partner.
The technical approach combines light from several weaker lasers, each operating on a different wavelength, into a single fibre. Weaker emitters fail less often during manufacturing, so a larger share passes testing, according to the company. The design targets up to 40 milliwatts of power per wavelength. That choice trades one powerful source for an array of smaller ones multiplexed together. For data centre operators, the promise is steadier yields and supply at a time when optical links must scale from telecom volumes to much larger quantities.
AI data centres increasingly move data between chips with light instead of copper wires, and each optical link needs a laser to supply that light. The existing supply chain was built for telecoms, a market that requires far fewer units. Chief executive Paul Marchal said the industry will soon need millions of these laser sources every year. Demand is therefore shifting from low-volume precision components to high-volume manufacturing. Photon Bridge positions its multi-wavelength design as an answer to that change in scale and production discipline.
What high-volume lasers mean for AI operators
For companies running or renting AI infrastructure, more manufacturable laser sources could ease expansion of high-bandwidth clusters. Stable supply of optical components affects lead times for networking upgrades and new capacity. A design with higher production yields can lower the risk of delays tied to a single laser supplier. Large operators gain another potential vendor for co-packaged optics and interconnect roadmaps. Smaller firms benefit indirectly if cloud providers can add capacity on schedule and keep performance consistent.
The limits sit in timing, qualification and unproven scale. Samples are expected only in 2027, with mass production preparation from 2028, so near-term purchasing decisions cannot rely on this product. Buyers should ask about test data at 40 milliwatts per wavelength, lifetime, supplier qualification and compatibility with their transceiver platforms. The announcement alone does not prove cost, reliability or customer acceptance. Competition is also active: Nvidia has put at least $6.5B into photonics firms since March, while CScale raised $145M and Volantis raised $88M for related optical technologies.
The marker to watch is delivery of first customer samples in 2027 and disclosure of testing partners. Names of chipmakers, yield figures and follow-on orders would show whether the multi-laser design transfers from lab to factory. Confirmed qualification by a major systems vendor would signal that supply for AI interconnects is broadening. Without those steps, the project remains a funded development plan rather than a supply option.
