Photon Bridge, a photonics start-up based in Eindhoven, has raised €10m to develop laser sources for the optical connections that link processors in artificial intelligence data centres, the company said today.

The round was co-led by two new investors, the Amadeus APEX Technology Fund and 360 Capital. Existing backers Innovation Industries, Faber and the Brabant Development Agency, known as BOM, also increased their investment.

The company said the money would be used to accelerate product development and customer qualification, deliver its first samples in 2027 and prepare for volume manufacturing from 2028. It also plans to expand its team in Eindhoven as it moves towards commercialisation.

Photon Bridge cited a forecast from Omdia, the research firm, that more than 20 million AI accelerators for data centres will ship each year by 2028. The company said that as AI clusters grow from thousands to hundreds of thousands of processors, moving data between chips is becoming one of the fundamental constraints on performance.

Optical connections can move larger volumes of data faster and more efficiently than electrical links, according to the company, and are moving closer to the processors themselves. That shift is accelerating with co-packaged optics, which places optical connectivity directly alongside the chips. Photon Bridge said the change creates a need for a far greater supply of the lasers that generate the light.

Goldman Sachs Research has forecast that the addressable market for optical networking will grow ninefold to $154bn by 2028, driven by the build-out of AI infrastructure and rising computing power per rack.

According to Photon Bridge, today’s supply chain for high-power lasers was largely built for telecoms and subsea networks, where volumes are far lower. The company said AI could turn the laser from a specialist photonics component into something that has to be manufactured at semiconductor scale.

Rather than generating the required optical power from a small number of increasingly powerful lasers, Photon Bridge combines and multiplexes light from multiple lower-power semiconductor lasers.

The company said lower power densities make individual lasers less sensitive to manufacturing defects, which improves production yield and allows the architecture to scale more efficiently as volumes rise. The platform is designed to deliver high wavelength counts with up to 40 mW of optical power per wavelength per fibre, and to be produced across multiple indium phosphide foundries.

Paul Marchal, chief executive of Photon Bridge, said: “AI is fundamentally changing the way computing infrastructure is built. We are moving towards enormous systems in which hundreds of thousands of processors have to operate together, and moving data between them is becoming one of the defining constraints on performance.”

He added: “That means optics has to move closer to the compute. But it also creates an entirely different manufacturing challenge. The industry will soon need millions of these laser sources every year. Instead of making individual lasers increasingly powerful and difficult to manufacture, we move that scaling challenge into the architecture.”

Ion Hauer, investment partner at APEX Ventures, said: “Every new wave of compute creates new bottlenecks further down the supply chain, and lasers are the next one. Photon Bridge is addressing this with a technology that combines deep photonics expertise with a credible path towards manufacturing at very large scale.”

Cesare Maifredi, partner at 360 Capital, said: “AI is turning lasers from a telecom component into a semiconductor-scale product. Once you start talking about millions of units a year, manufacturing yield becomes fundamental. Photon Bridge has designed its architecture from the beginning around the volumes this market is going to require.”

APEX Ventures launched the Amadeus APEX Technology Fund with Amadeus Capital Partners in early 2023. 360 Capital, based in Paris and Milan, says it manages more than €750m.

The funding comes as attention turns to the physical limits on AI build-out. Last month a Tony Blair Institute report argued that the supply of chips and energy systems, rather than code, caps how much AI can be built. In the UK, the BBC reported on Monday that Microsoft had pulled out of Nscale’s planned £2bn Loughton data centre in Essex, where the grid connection has been pushed back to the 2030s.

Read the full article →