Europe Can Make Photonic Chips. Who Will Buy Them?
European companies are beginning to design systems around photonics. The next industrial test is whether they will commit to buying it at scale.
Grenoble has changed the terms of Europe’s photonics debate. STMicroelectronics is producing silicon photonics on 300 mm wafers, and its expansion plans are backed by long-term capacity reservations. The question is no longer whether Europe can manufacture a photonic chip for the AI economy. It is whether enough European customers will build lasting businesses around that capacity.
For now, the clearest route to scale runs through the global AI market. ST describes its high-volume PIC100 customers as leading hyperscalers; Amazon Web Services has publicly identified itself as a collaborator in developing the technology. Those relationships bring production in France the volumes and commercial discipline it needs. They also give large customers substantial influence over the technology’s direction.
This is Europe’s industrial paradox. Its manufacturers need to sell where demand exists today. European system makers may need more time to develop products they can order at scale. If those schedules diverge for too long, Europe could become an excellent place to make chips for systems specified elsewhere.
If those schedules diverge for too long, Europe could become an excellent place to make chips for systems specified elsewhere.
The European customer is beginning to emerge. The harder question is when that customer will be ready to buy.
Customers before orders
The STARLight programme offers a useful place to look. Led by ST, it brings manufacturers, researchers and application developers together around 300 mm silicon photonics. Its work through 2028 is directed at data centres, AI, telecommunications and automotive applications, as well as the manufacturing chain needed to supply them.
Several partners have defined uses for the technology. Ericsson is exploring an integrated optical switch for mobile networks and a radio-over-fibre approach that could move power-intensive processing away from antenna units. Thales is developing sensing and signal-processing functions. German optical module maker Sicoya has linked ST’s platform to its future product line; the companies have presented a PIC100-based 1.6T transceiver demonstrator.
A demonstrator establishes a technical route. A production order establishes that a customer expects to use it repeatedly.
These companies matter because they can help decide what the chip must do. They can set requirements for a network device, a sensor or an optical module, then test whether a photonic platform meets them in a working system. That is a more consequential role than simply appearing on a consortium’s list of partners.
It is still distinct from committing to recurring purchases. A demonstrator establishes a technical route. A production order establishes that a customer expects to use it repeatedly, at a price and quality the supplier can sustain. Europe has visible evidence of companies working toward the first step. Public evidence of large European commitments at the second step is harder to find.
The market is wider than the AI cluster
AI data centres attract attention because their appetite for optical connections is immediate and enormous. But a European photonics market need not depend on producing a smaller version of the American hyperscaler business.
SteerLight, another STARLight partner, is developing compact LiDAR sensors for vehicle applications. The French company says secure component supply is important to its plans for larger-scale production. A successful product could connect a European photonics platform to a European sensor maker and, potentially, to vehicle manufacturers. Yet automotive adoption requires testing, qualification and product decisions beyond a successful chip demonstration. It cannot be assumed to fill a fab’s order book next year.
Telecommunications offers another opening, with an important distinction. Orange Belgium has awarded Nokia a multi-year contract to modernise its optical transport network. The contract shows that a European operator is spending on optical infrastructure. It does not establish that the network will use photonic integrated circuits manufactured by ST, SMART Photonics or another European foundry. Network equipment procurement and chip procurement sit at different points in the chain.
The same caution applies to public computing. EuroHPC has signed a €387.8 million procurement contract for the LUMI-AI supercomputer in Finland. That creates a substantial European buyer for advanced AI infrastructure, but the published contract does not identify a purchase of European photonic chips. The prospect lies in connecting such system-level demand to component design and sourcing decisions early enough to matter.
The clock is already running
Europe is building ways for prospective customers and suppliers to work together. The photonixFAB project connects X-FAB, SMART Photonics, PHIX, Luceda and other production partners with application developers including Nokia and Thales. It covers design, manufacturing, packaging and demonstrators. Its own reporting describes early customer engagement while recognising that the project remains at an early stage.
PIXEurope addresses a related gap by giving companies shared access to design, prototyping, integration and testing. More than 100 companies and research organisations had engaged with the pilot line through technical discussions and exploratory collaborations by May 2026. Its full range of services, including multi-project wafer runs, is expected by May 2028. Interest is growing. The transition from exploration to repeatable industrial demand still has to happen.
The commercial market will not wait for every European pilot line to reach full capacity.
ST says its planned production increase is supported by long-term reservations from customers it describes as leading hyperscalers. SMART Photonics is recruiting a sales director on the American west coast to pursue AI, datacom and optical-networking business. These are sensible decisions for manufacturers that need customers now. They make the timing question unavoidable for Europe: can its own system makers move from development to purchasing while they can still help shape the platforms being scaled?
An anchor customer is a commitment
European policy is beginning to recognise the demand problem explicitly. The Commission’s proposed Chips Act 2.0 includes measures to connect chipmakers with user industries, stimulate demand through innovation procurement and link semiconductor production to the growth of European data centres and AI infrastructure. A proposal is not an order, but it marks a useful shift in emphasis: building capacity and building its market have to proceed together.
For photonics, an anchor customer would have to do more than express interest in a European supply chain. It would help define a product, test it in a system and commit to a credible route toward repeated purchases. That commitment would allow suppliers to plan capacity and improve yields against a real application. It would give designers and packaging specialists a reason to invest around the same product.
There will be no single customer for every photonic chip. Telecom networks, automotive sensors and AI clusters have different requirements and buying cycles. The practical task is to turn promising European applications into specific commercial relationships, one market at a time.
Grenoble demonstrates that Europe can hold a place on the wafer. Its next test is whether European customers can help decide what belongs on it.
What would change the picture? An announcement naming a European system maker, the European photonics platform it will use, the product in which it will appear and a credible path to volume. That would tell us more about industrial progress than another list of consortium members.
Europe’s photonics customers are beginning to appear in the design process. Ericsson, Thales, Sicoya and SteerLight show that companies here have uses for the technology of their own. Public computing and telecom spending show that Europe can also purchase advanced systems. The missing proof is a visible bridge between those two facts: European products, built with European photonic chips, bought in volumes that endure.
Grenoble demonstrates that Europe can hold a place on the wafer. Its next test is whether European customers can help decide what belongs on it.
Perspective: Europe’s photonics market is waking up as a designer of systems. Its arrival as a large, committed buyer of chips has yet to be demonstrated.
Image credit
AI-generated illustration for Altair Media.
Caption
European photonics production is growing. The next test is whether European companies will turn chip designs into products they buy at scale.
