OpenLight has fresh funding, production customers, named ecosystem partners, and a 400-gigabit modulator roadmap aimed at optics for AI.
OpenLight has raised $50 million in Series A-1 funding, bringing its total funding to $84 million.
The round, led by Matter Venture Partners, is intended to accelerate deployment of the OpenLight’s heterogeneous silicon photonics platform and expand customer adoption across a range of segments: AI infrastructure, automotive, industrial sensing, medical, and quantum computing applications.
The timing is significant for OpenLight as it looks to turn its indium-phosphide-on-silicon photonics technology into a production platform used by many customers.
OpenLight, which positions itself as a provider of heterogeneously integrated silicon photonics platforms, is moving from prototype activity to production programmes.
“I call it a beachhead strategy,” says Adam Carter, CEO of OpenLight. “You land on the beach and then you fan out.”
Customers
Carter says well-known players are now looking at its technology that will lead to work with some “interesting companies”.
One company already named is NewPhotonics, the Israeli start-up that has placed volume production orders using OpenLight’s process at foundry Tower Semiconductor.
“When we go to customers, the first thing they ask is: who’s in production?” he says. “We’ve been very open and very honest, and we say: we’re prototyping.” Now that the work with NewPhotonics is publicly announced, customers can understand that there are now customers placing orders using its process at Tower Semiconductor.
NewPhotonics was one of OpenLight’s earliest customers, soon after OpenLight was spun-out from Juniper Networks.
NewPhotonics used OpenLight’s process design kit (PDK), hosted at Tower, to design its optical engine photonic integrated circuits (PICs). OpenLight provided the active and passive building blocks; NewPhotonics did the PIC design.
“They took all our actives, they took the passives, and they designed the PIC,” says Carter. “They were the first company that has given us PIC reliability data based on our PDK and our components.”
That matters because OpenLight’s business model depends on repeatable foundry use. Customers license certain components and OpenLight receives royalties on production wafers shipped from Tower. OpenLight is not trying to sell one customer one chip; it is trying to make its integrated laser, modulator, amplifier, and detector platform a foundry option.
The company model is not one-size-fits-all. NewPhotonics needed relatively little help. Oriole, the University College London spin-out developing optical switching infrastructure for AI workloads, worked with OpenLight differently.
“Oriole provides the system specifications, and we work closely with them on the PIC design and tape-out process,” says Carter. “Then we manage their runs through Tower for the development phase.”
That flexibility is important. Some customers want PDK access and independence. Others need design services. OpenLight supports both.
The wider setting for all this activity is AI and the crucial role of optics in enabled AI computing. Carter says the recent OFC show, held in Los Angeles, had the feel of 1999 in terms of activity and supplier optimism.
In 1999 the optimism was fueled by the dotcom era, but now there is a crucial difference, says Carter. Then, companies added capacity based on forecasts that failed to materialise. This time with AI, large customers such as Nvidia, with its investments in Coherent and Lumentum, are putting money into the supply chain.
That distinction explains why Carter is cautious about bubble comparisons. “AI is not going to go off a cliff like the market did in 2000,” he says. “More likely, the growth rate will change.”
Technical focus
OpenLight’s technical focus is equally clear: 400-gigabit-per-lane modulation and future 3.2-terabit optics.
OpenLight first announced a 400-gigabit optical modulator on Tower’s process in 2025 and has spent the past year improving it. The company diverted some beta 1.6-terabit wafers and applied process changes to its 200-gigabit modulator to produce a 400-gigabit version.
“We shipped 3.2-terabit DR8 prototypes to about five transceiver companies in January,” says Carter. “You’re going to say that’s way too early, but you have to be early at the component level for the next generation of transceivers.”
At OFC, OpenLight demonstrated its 400-gigabit-per-lane electro-absorption modulators in combination with MACOM’s 400-gigabit modulator driver technology.. Carter says its device showed PAM4 signalling with an extinction ratio of 4.6dB and a 3dB bandwidth of roughly 85GHz using a 2V differential driver. OpenLight is confident the bandwidth will be improved. And using a single-ended drive, a bandwidth of 110GHz was achieved.
OpenLight’s other claim about its 400-gigabit modulator is its size. OpenLight’s indium-phosphide modulator is about 100 microns long. Carter argues this gives indium phosphide a density advantage for future co-packaged optics, where modulator arrays may become very large and package dimensions tightly constrained.
Thin-film lithium niobate, says Carter, has excellent performance and is a little bit ahead in some respects. But he questions whether it can deliver the density required for wide-channel co-packaged optics.
Micro-ring modulators, which OpenLight also has in its library, are compact and attractive for multi-wavelength slow-and-wide architectures, but bring with them thermal-stability challenges.
Carter’s conclusion is not that one modulator technology wins everything. The market may be large enough for several approaches. But he is clear where OpenLight wants to play: “Indium phosphide at 400 gigabit is definitely a way to go.”
That said, Carter stresses that OpenLight is not just an indium phosphide company. “If a customer wants a specific technology and it can be incorporated into a silicon photonics process, we will do something to have that in our PDK,” he says. “So customers can have a choice.”
OpenLight now has fresh funding, production customers, named ecosystem partners, and a 400-gigabit modulator roadmap aimed at optics for AI.
But Carter remains circumspect. The company has landed on the beachhead. Now comes the fanning-out stage.