The problem: Chips are too complex to see

Modern chips stack logic and memory in three dimensions, with features just a few nanometres apart. Optical metrology tools—the microscopes that check each layer—can no longer resolve buried structures. That creates a yield problem: if you can't confirm a layer printed correctly before adding the next, you're building blind. In a business where a marginal yield gain is worth billions, that blind spot hurts.

Invisix's answer: Soft X-rays from a Nobel-winning physics trick

Invisix, an Eindhoven startup spun out of ASML, has raised €20 million to fix that blind spot. The oversubscribed seed round drew Hitachi Ventures, Transition Ventures, imec.xpand, Doosan Investment Co., and a tier-1 semiconductor manufacturer—likely Samsung, according to earlier reporting, as it works on yields for its 2nm process.

Their system uses high-harmonic generation—the physics behind the 2023 Nobel Prize in Physics awarded to Pierre Agostini, Ferenc Krausz, and Anne L'Huillier. A short-pulse laser excites noble-gas atoms until they emit soft X-rays across many wavelengths at once. Unlike a single-colour laser, this broad spectrum produces a richer 3D signal. The company pairs that with reconstruction algorithms and machine learning to rebuild a device's internal structure without destroying it.

Retired risk: A decade of ASML R&D

What makes this bet unusual for a seed-stage hardware company is how much risk is already retired. Invisix licensed a substantial soft X-ray technology package developed over more than a decade inside ASML. Its team is stocked with veterans of that programme, plus COO Roald Dogge, formerly of contract manufacturer NTS. The company says it demonstrated the approach publicly in 2023 with Intel and imec, measuring features in gate-all-around (GAA) transistors—among the hardest targets for existing metrology.

Why it matters for developers

For developers working on chip design, EDA tools, or semiconductor manufacturing software, this means better data for process control. If Invisix's system works at volume, it could reduce the need for destructive testing (slicing wafers) and speed up process development for new nodes. The reconstruction algorithms and ML models are likely to be integrated into existing metrology workflows, potentially via APIs or data exchange formats. For now, the company is building its first shippable system and running customer demos from a new Eindhoven cleanroom.

The wider context: Europe's chip supply chain

The Netherlands is home to ASML, and a cluster of Dutch and Belgian deep-tech funds—including imec.xpand—now backs the startups forming around it. Invisix has the testbench, the pedigree, and the cleanroom. The test ahead is the unforgiving one for any tool company: turning a demonstration into a system a fab will run at volume.

Technical deep-dive: Soft X-ray metrology

Soft X-rays have wavelengths in the range of 1–10 nm, shorter than extreme ultraviolet (EUV) light used in lithography (13.5 nm). This allows them to penetrate multiple layers of a chip and interact with buried structures. Invisix's system uses a laser-driven plasma source to generate a broadband spectrum, which is then directed at the chip. The scattered X-rays are collected and analysed by algorithms that reconstruct the 3D geometry. This is similar to computed tomography (CT) scans in medicine, but at nanometre resolution.

The company demonstrated the approach in 2023 with Intel and imec on GAA transistors. GAA transistors have horizontal nanosheets surrounded by gates, making them extremely hard to inspect with conventional methods. Invisix's system reportedly resolved features down to sub-2nm, though exact numbers are not public.

Competitive landscape

Other metrology approaches include scanning electron microscopy (SEM), transmission electron microscopy (TEM), and atomic force microscopy (AFM), but all are slow, destructive, or limited to surface inspection. Optical critical dimension (OCD) scatterometry is faster but struggles with 3D geometries. Invisix's soft X-ray approach aims to combine speed, non-destructiveness, and 3D capability.

Next steps

Invisix will use the €20M to grow its team, build its first shippable system, and run customer demonstrations. The company's success will depend on whether it can integrate its tool into high-volume manufacturing fabs. For developers, the key is to watch for integration details—likely in the form of software SDKs or data formats—that will allow process control systems to consume the metrology data.

References