Between AI frenzy and cleanroom precision: the photonics plant

On the southern edge of the High Tech Campus in Eindhoven, the concrete skeleton of what is set to become the epicentre of European photonics is taking shape. While BAM construction workers manoeuvre around the scaffolding and the façade's contours become visible, a different dynamic is unfolding behind the scenes. Two years ago, the market still held healthy scepticism about the usefulness of a large-scale industrial pilot line. Today, customers are practically queuing at the gates.

“When we started this project two years ago, many people were wondering: can you actually fill that factory, and with what kinds of applications?” says Boudewijn Docter, who is closely involved in developing the new facility on behalf of TNO. “That sentiment has now completely reversed. Nobody doubts anymore that we will be at full capacity as soon as the facility becomes operational.”

The driver behind this sudden fever is easy to identify: the global explosion in investment in artificial intelligence and data centres. Conventional electronic chips are approaching their thermal and physical limits; photonic chips, which transport data using particles of light rather than electrons, offer much-needed gains in speed and energy efficiency. As a result, demand for optical chips is surging across the supply chain, with partner SMART Photonics playing a significant role.

No shortcuts at the micrometre scale

But anyone who thinks that building a chip factory can go faster simply by stepping on the accelerator is mistaken. Between the wishes of AI customers hungry for capacity and the unforgiving reality of nanotechnology lies a process that takes years. Docter feels the weight of the expectations that have built up, but remains level-headed: “There is a great deal of pressure on us to become operational sooner. Customers are waiting; they urgently need production capacity. But at the same time, we cannot and do not want to take shortcuts. We have to look at what is realistic.”

That difference lies in the project's very nature. This will not be an academic playground where failure is part of the experiment. “In an R&D setting, you can sit down with a customer, develop a product together and share the risk involved in how quickly you can get something operational,” Docter explains. “But this is a real production facility. Essentially, the product design must be finished, signed off and ready to scale. You cannot then show up with a production line that is still only half built or that you are not certain can handle the process. Customers simply will not take that risk.”

According to Docter, customers will only truly scale up commercially once you can demonstrate a proven track record. “That requires hard statistics and process stability. You cannot build those up overnight.” TNO must therefore demonstrate not only that the processes work, but also build up evidence of their reliability and reproducibility. Only then, according to Docter, customers will dare to entrust their large production volumes to the line.

2027, 2028, 2029: the strict roadmap

The structural construction is proceeding right on schedule. If everything stays on track, the building will be formally handed over in May next year. From that point onwards, the precision work inside will really begin:

• End of next year (2027): All core equipment must have been installed and connected inside the cleanroom. Most of the orders, subject to strict European procurement procedures and delivery times of up to twelve months, have now been placed. These include lithography systems from ASML, growth systems from AIXTRON, and etching equipment from KLA and Oxford Instruments. Each machine must also be running and validated on the basic processes by then.
• 2028: The year of integration. Hundreds of individual processing steps must link into a single, seamless process flow. The pilot line will switch from the current 4-inch standard to larger 6-inch indium phosphide (InP) wafers, which should drastically reduce the cost per chip.
• 2029: The first actual projects for consortium partners within the PIXEurope programme.
• 2030: The phase of full-scale commercial operation.

A significant share of the equipment for the new plant comes from suppliers that are also well known in the traditional semiconductor industry. That is no coincidence. According to Docter, the new factory should make as much use as possible of existing industrial equipment. “In principle, they are all standard semiconductor tools, but some will have to be modified.”

That applies, for example, to ASML’s lithography machines. They must work with indium phosphide instead of the silicon wafers for which many chipmaking machines were originally developed. In addition, many modern systems are designed for wafers measuring 8 inches or more. “And we are going to use them for 6-inch wafers. That sometimes requires a minor modification.”

Other equipment is already suitable for so-called compound semiconductors such as indium phosphide, gallium arsenide, gallium nitride and silicon carbide. Such materials are also used in applications including power electronics, fast chargers and high-speed radio communications.

The shadow of the next multibillion-euro factory

Perhaps most tellingly, discussions are already underway about the next factory. According to earlier plans, the current pilot line could eventually support a production capacity of approximately 20,000 6-inch wafers per year. To serve a truly international mass market, it will ultimately be necessary to scale up again to a factory with at least five times that capacity.

So while TNO, SMART Photonics, High Tech Campus Eindhoven and Eindhoven University of Technology are pulling out all the stops to prepare HTC 12, the debate about the next phase is already raging beyond its walls. Eindhoven mayor Jeroen Dijsselbloem and former ASML CEO Peter Wennink recently warned that €2 billion will have to be raised within two years for a fully fledged industrial megafactory: a facility ten times larger than the current pilot line. After all, foreign players in the US and Asia are not standing still.

Docter acknowledges the need to look ahead: “Just look at how complex permits, grid connections and locations have become these days. Here at High Tech Campus Eindhoven, the facility fits within the zoning plan and the existing environmental parameters, but for an even larger next step, you have to start planning now.”

Still, the TNO programme manager refuses to be swept up by the debate over billions in The Hague or Brussels. In his view, the best way to secure a leading European position is to first deliver the factory currently under construction flawlessly. “Our focus is on the current project: getting it up and running, validating it and delivering.”