小批量大杠杆：欧洲在先进芯片制造中的隐形角色

At the Fraunhofer Institute for Silicon Technology (ISIT) in northern Germany, engineers manufacture a very small number of specialized semiconductor chips. By the standards of an industry accustomed to producing millions of chips, the output is negligible. Yet these components help operate equipment used by Intel and TSMC to manufacture some of the world’s most advanced semiconductors.

The customer is IMS Nanofabrication, an Austrian company that builds multibeam electron-beam systems for rapidly writing photomasks used in EUV lithography at technology nodes of 5 nm and below. Fraunhofer supplies a critical component of those machines: a chip containing hundreds of thousands of individually controlled structures that allow each electron beam to be controlled independently.

The relationship illustrates a little-recognized part of the semiconductor value chain. While governments compete to attract advanced fabs, Europe already supplies specialized technologies on which those fabs depend.

“We’re operating in a niche that conventional semiconductor foundries aren’t really set up to serve,” Fabian Lofink, deputy director of Fraunhofer ISIT, told EE Times. “The volumes are very small. The technology is extremely complex. At these production levels, you can’t rely on statistical process control the way high-volume fabs do, so engineers need to be closely involved in manufacturing every day. “

By Hynetek Semiconductor Co., Ltd.  10.15.2026

By Ayar Labs  10.14.2026

By ALI  10.14.2026

From a thousand beams to hundreds of thousands

The collaboration began in 2005 through publicly funded research. As semiconductor features shrank, the corresponding lithography masks required increasingly dense patterns, driving up electron-beam writing times. Multibeam technology offered a solution by allowing many electron beams to write in parallel.

“As feature sizes shrink, you need to write far more data onto the mask,” Lofink explained. “With a conventional single-beam writer, that directly translates into longer write times. The idea behind multibeam technology was to parallelize the process, enabling high-throughput mask writing even for the most advanced technology nodes.”

The first demonstrator measured one square millimeter and contained 1,000 cells. By 2012, Fraunhofer and partners had developed a chip with approximately 262,000 individually controlled cells.

Fraunhofer does not manufacture the underlying CMOS circuitry. Instead, the chips are fabricated using advanced CMOS technology with feature sizes below 50 nm to achieve the data rates required for multibeam operation. Fraunhofer then adds the structures that enable individual electron beams to be controlled.

This post-CMOS processing requires electroplated gold electrodes, chosen for their high conductivity and ability to support rapid beam control.

“Gold is normally something you don’t want anywhere near a CMOS fabrication line because of contamination,” Lofink said. “But in our application, even the slightest charging effects can distort the electron beam trajectories. Gold provides both the excellent conductivity needed for high-speed beam control and a reliable path for dissipating charge. Our expertise is in combining these materials after the CMOS process without damaging the electronics.”

Some gold structures approach 50 micrometers in height, far beyond conventional planar semiconductor dimensions.

Around 2015, IMS began selling mask-writing systems incorporating the second-generation TROM2 chip to produce photomasks for the 7-nm technology nodes. Fraunhofer continues to manufacture that component.

Too specialized for conventional foundries

The biggest question is: Why does IMS continue to buy from a research organization rather than a commercial foundry?

“Conventional foundries are looking for business cases involving at least 10,000 wafers a year,” Lofink said. “We’re operating at volumes roughly two orders of magnitude below that threshold.  That’s a completely different manufacturing environment.”

The Fraunhofer process also involves approximately 15 lithography levels beyond the CMOS fabrication, specialized materials, and demanding quality requirements.

“High complexity, low volume—that’s our niche,” Lofink said. “You cannot manufacture these products simply by putting operators in front of machines and asking them to push buttons. You need engineers working with the process every day to maintain the required quality.”

High-volume manufacturers use extensive production data to identify variations and stabilize processes. Very small production runs provide fewer opportunities to accumulate such data, increasing the importance of direct engineering supervision.

Fraunhofer also combines experimental research with the ability to supply components reliably. “What differentiates us from a university is that we can take research into an industrial environment,” Lofink said. “We can deliver the same quality repeatedly, and we can deliver it when the customer needs it.”

A small supplier with strong influence

The arrangement has created a supply chain dependency. Fraunhofer remains the key source of TROM2, and transferring production elsewhere would require substantial development.

“If our production stopped, IMS would have a serious problem,” Lofink said. “They have some stock, of course, but you cannot simply move this technology to another supplier. Developing the process takes years.”

For the latest version, TROM3, which is the third generation of the chip, IMS has expanded its manufacturing ecosystem. According to Lofink, it contains approximately 590,000 control cells and supports higher data rates.

The importance of the chips extends far beyond their production volume. IMS’s systems write masks for advanced semiconductor manufacturing, including EUV processes. Lofink identifies Intel and TSMC as major users.

The chips themselves use comparatively mature CMOS technology. “The CMOS in our chip might use a 100-nm or 50-nm process, but the equipment it goes into is used to make masks for 5-nm or 2-nm technology,” Lofink explained.

Extending the model across Europe

For Dirk Schumann, director of the APECS pilot line at the Research Fab Microelectronics Germany (FMD), the IMS relationship illustrates an industrial model that Europe can expand. “If you compare the price of the components we supply with the industrial business they enable, the leverage can be a factor of 50 to 100,” he told EE Times.

The estimate reflects his assessment of the potential downstream value of specialized components, rather than a measured return for the IMS collaboration. Schumann distinguished between the requirements of high-volume chipmakers and those of equipment manufacturers, medical technology companies, and other specialized customers.

“If you’re talking about a company such as Intel or TSMC, they may come to a pilot line for early development, but once they move toward volume manufacturing, they need a different infrastructure,” he said. “For an equipment manufacturer, it’s different. They may need only a limited number of very advanced components, and we can continue supplying those components.”

FMD operates through 13 Fraunhofer institutes and two Leibniz institutes, offering capabilities from technology development through prototyping and small-volume manufacturing. “Our customers don’t need to own the entire infrastructure by themselves,” Schumann said. “We can take a company from an initial idea through technology development, design, prototyping, and even small-volume manufacturing.”

The IMS collaboration predates APECS by many years. Although it didn’t create that relationship, APECS seeks to extend the underlying model through additional capabilities and European partners. Its distributed network aims to give industrial customers access to specialized facilities through a single point of contact, spanning materials, semiconductor processing, and advanced packaging.

Schumann emphasized that developing new technologies is only part of Fraunhofer’s mission. “The real difference comes when we can turn those technologies into industrial value by supplying components that companies can use in their products,” he said.

From technical capability to industrial delivery

APECS is developing 11 demonstrators across four technology areas: high-performance computing, sensors, optical technologies, and RF applications. Beyond proving the technologies themselves, the projects are testing whether geographically dispersed research facilities can operate as a coordinated industrial supplier.

“Technically, moving wafers or chiplets between different institutes isn’t necessarily the biggest challenge,” Schumann said. “The interesting part is making the whole system work as a business: preparing quotations, planning production, coordinating the different facilities, and delivering what the customer has ordered.”

Industrial interest is emerging. According to Schumann, four demonstrators have attracted customers seeking evaluation hardware. “The total number of project requests at APECS, primarily from industrial customers, has already exceeded 200,” he said. How many will lead to sustained commercial production, and when, remains uncertain.

The Fraunhofer–IMS relationship demonstrates what that transition requires: not simply manufacturing a specialized component, but delivering it consistently, reliably, and on schedule. APECS aims to replicate that model across a much larger European network.

Europe’s contribution to advanced chipmaking cannot be measured by production volume alone. Fraunhofer manufactures the specialized chips it supplies to IMS in tiny quantities, yet they play a critical role in equipment used by some of the world’s largest semiconductor manufacturers.

Read also:

Europe’s Advanced Packaging Gap: Fraunhofer, Swissbit Target the Missing Middle

Fraunhofer IPMS Eyes India For Contract Research and Semiconductor Partnerships

Europe’s Strategic Bet: Heterogeneous Integration on 300-mm Wafers

Silicon Saxony Shows Promise, Limits of Europe’s Chips Act

CHIP MANUFACTURING, E-BEAM, EUV LITHOGRAPHY, PHOTOMASKS, SEMICONDUCTORS

FRAUNHOFER INSTITUTE FOR SILICON TECHNOLOGY (ISIT), IMS NANOFABRICATION, INTEL, TSMC