Custom chips as easy to iterate as PCBs, that's xSilica's mission

At the June edition of Gerard & Anton’s Demos Pitches & Drinks, Jurgen van den Berg argued that Europe’s chip autonomy may not begin with ever-smaller transistors, but with faster, cheaper experimentation.

Every hardware team knows the silence Jurgen van den Berg described on stage at the TU Eindhoven campus. "The product is growing. The system is becoming too complex. Power consumption is creeping upward. Firmware starts to feel heavy." And then, sooner or later, someone in the room says the sentence that changes everything: “Maybe we should build an ASIC.”

Van den Berg paused for effect during his pitch. “And then the room goes quiet,” he said. “Not because it’s a bad idea, but because everyone knows what it means.”

What it means, in today’s semiconductor industry, is time, money and risk. Months of waiting. Millions in investment. A development path in which a company may only learn whether the custom chip really works after a major capital decision has already been made. “Most teams can’t afford to go around twice,” Van den Berg said.

With his company xSilica, he wants to change that logic. Not by joining the global race toward the smallest possible transistor, but by moving in the opposite direction: toward mature-node silicon, smaller fabs, integrated tooling and rapid iteration.

The missing feedback loop

Van den Berg started from a simple observation. In almost every other part of hardware development, iteration has become fast and affordable. “Nowadays you can spin a new PCB, a circuit board, in two weeks,” he said. “You can 3D print a new enclosure overnight.”

That speed has changed how hardware teams work. Engineers can test assumptions, adjust designs, learn from prototypes and make new versions quickly. But custom silicon remains different. “Somehow custom silicon is still a one-shot capital event,” Van den Berg said.

That has consequences far beyond chip design itself. When ASIC development is too expensive or too slow to try early, product teams stay longer at the PCB level. They use off-the-shelf components. They add complexity around the chip rather than integrating functionality into it. The result can be bigger systems, higher energy use and software workarounds that make products less elegant than they could be.

For Van den Berg, the problem is no longer mainly physical. “It’s not really a physics problem anymore,” he said. “It’s more of an industry structure problem.”

Three verticals, one rigid system

To explain that structure, Van den Berg reduced the semiconductor world to three verticals. First, there is chip design software: EDA, or electronic design automation. Second, there are equipment manufacturers, with ASML as the most obvious example for an Eindhoven audience. Third, there are the fabs: the highly specialised factories that combine chemicals, equipment and recipes to manufacture chips.

Together, these verticals have achieved something extraordinary. “They’ve worked together for decades to build the smallest transistor at the largest scale,” Van den Berg said. That focus created the modern semiconductor industry, but it also created rigidity. The whole system was optimised for scale, density and extreme performance, not for fast learning cycles.

“No one’s ever thought of how to do this iteratively,” he said.

That is where xSilica comes in. Van den Berg describes the company as “one platform, fully integrated,” combining simplified EDA software, a proprietary chemical recipe and a fabrication recipe into a single automated process. The goal is not to replace the high-end chip industry. The goal is to create a different entry point: one designed specifically for rapid iteration of custom silicon.

A container-sized fab

There was a deliberate provocation in Van den Berg’s pitch. “If you want to build the smallest transistor, you have to build the biggest fab,” he said. “I think we should run in the other direction.”

Instead of chasing the most advanced nodes, xSilica focuses on mature-node silicon: larger structures, made with existing equipment, for applications where speed of development, cost and local availability matter more than ultimate transistor density. Van den Berg asked the audience to imagine “a container-sized fab here in Eindhoven or in Berlin or in Rome or in Paris.”

That image matters. Europe’s discussion about chip sovereignty often starts with multi-billion-euro megafabs. xSilica suggests another layer of autonomy: local or regional facilities that make custom silicon accessible to more companies, especially those that cannot justify a conventional ASIC route in the early stages of product development.

For Van den Berg, strategic autonomy is not only about producing the most advanced chips. It is also about giving European product teams the ability to experiment, learn and build without being locked into a slow and expensive global pipeline.

Why now?

Van den Berg gave three reasons why he believes the moment is right. The first is technological: automation and AI-assisted machine controls could make smaller, more integrated fabrication processes easier to run. The second is economic: demand for mature-node silicon is high, driven by industrial systems, automotive applications, sensors, power electronics and countless connected devices that do not need the world’s smallest transistors.

The third reason he delivered with a smile: “I think it’s Trump.”

Behind the joke is a serious point. Geopolitical pressure has changed how Europe thinks about technology. Dependencies that once seemed efficient now look vulnerable. Semiconductors are no longer just components in a supply chain; they are strategic infrastructure. “Europe’s finally figured out that we need to take care of ourselves,” Van den Berg said. He added that deep-tech platforms are becoming investable again.

That shift could create room for companies like xSilica. Not as a replacement for the semiconductor giants, but as a new category between software-like iteration and industrial-grade chip manufacturing.

Looking for a cleanroom

XSilica is still early. Van den Berg was open about that. “I’m currently very early stage, at the pre-seed stage,” he told the room. “I have my thesis ready and I’m looking for cleanroom access to validate it.”

It was a fitting ending for a Gerard & Anton pitch: ambitious, concrete and directed straight at the ecosystem. Sometimes it begins by making the next iteration smaller, faster and close enough to try again.