Quantum Computing Scaling Is a Packaging Problem, EV Group Argues

Quantum Computing Scaling Is a Packaging Problem, EV Group Argues

EV Group Says Quantum Computing Scaling Is a Packaging Problem

EV Group is positioning itself in quantum computing scaling through packaging rather than physics, applying wafer bonding, lithography and thin-wafer processing to the manufacturing problems that appear when a quantum architecture moves from a laboratory demonstration toward something built at volume. The Austrian company presents the work at SEMICON Europa in Munich this week, and its argument is that quantum computing is not one technology but a collection of fundamentally different approaches, each with its own materials and manufacturing requirements.

Why qubits need packaging at all

The case for advanced packaging in quantum computing starts with error correction. A qubit is fragile, and the error rates achievable today mean a fault-tolerant machine needs many physical qubits to represent one logical qubit. Estimates in the field vary by architecture and by assumption, but the direction is consistent: useful machines require far more qubits than any current design integrates.

That turns an integration problem into a manufacturing problem. If a system needs thousands of qubits, and each qubit needs control lines, readout and isolation from its neighbours, the wiring becomes as hard as the physics. A superconducting qubit operating at millikelvin temperatures cannot connect to room-temperature electronics by a wire per qubit beyond a certain count, which is why three-dimensional integration matters.

EVG’s framing is that the industry’s challenge is not simply increasing qubit count but finding ways to integrate them in increasingly complex chip architectures without compromising performance. That is a packaging statement, and it is correct as far as it goes, which is further than most capability claims in this space.

Why packaging became central

Packaging has been the unglamorous half of semiconductor progress for two decades, and quantum makes it central. In classical computing, advanced packaging became strategic when Moore’s Law slowed and the industry could no longer rely on transistor shrinks alone; the answer was to stack and to combine chiplets. Quantum faces the same problem earlier and harder, because the devices are more sensitive and the materials more varied.

That places the companies supplying bonding and lithography tools in a position that looks peripheral and is not. A quantum computer is a system, and its performance depends on how well the parts are joined as much as on how good the parts are. EVG’s bet is that this makes packaging expertise a durable position rather than a service to a speculative industry.

The per-architecture problem

The most useful part of the release is the mapping between process technology and qubit type, because it explains why a single manufacturing solution does not exist.

For superconducting devices, oxide-free bonding supports superconducting interconnects and three-dimensional device integration. The requirement is that the bond itself introduces neither resistance nor an oxide layer that disrupts superconductivity, a demanding constraint on surface preparation and vacuum conditions.

For photonic devices, fusion bonding and oxide-free bonding enable integration of III-V materials and high-Pockels-coefficient materials onto waveguides. That matters because silicon is a poor electro-optic material, and efficient modulation requires a material with a stronger Pockels effect bonded onto silicon photonics.

For spin qubits, hybrid bonding supports ultra-high-density interconnects, which is a different problem again: spin qubits in silicon need dense electrical access rather than optical or superconducting connections.

For trapped ions, EVG says it has worked with researchers on low-temperature metal bonding processes enabling integration of ion-trap and photonic chiplets made from different substrate materials. Trapped-ion systems need vacuum, optics and electrical control in one package, and the chiplets doing those jobs come from different material systems.

Two of the six approaches named in the taxonomy go unmapped in the process discussion. Neutral-atom systems trap atoms in optical tweezers and need free-space optics rather than bonded waveguides, and nitrogen-vacancy diamond qubits operate in a material difficult to integrate with silicon processing at all. Neither appears in the mapping, which may reflect where EVG sees near-term demand rather than a gap in the portfolio, but the release does not say which it is.

The industry asymmetry

A broader industry pattern is worth naming. Quantum hardware companies tend to be small and specialised, developing one architecture and one materials set. The equipment suppliers they depend on are large, established and sell into classical semiconductor markets.

That asymmetry shapes what a partnership looks like. A quantum startup cannot develop its own bonding process from scratch; it needs a tool vendor’s process knowledge and a facility where trials can run. An equipment company cannot develop a quantum-specific product until enough customers need the same thing. The competence centre model, where a vendor hosts process development for multiple customers, is how both sides resolve that. EVG operates two, and the CHAMP-ION pilot line is a third variant of the same idea with public funding behind it.

What EVG actually offers

The portfolio spans fusion and hybrid bonding, high-vacuum oxide-free bonding, digital and nanoimprint lithography, and thin-wafer processing, alongside temporary bonding and debonding. Two named competence centres, the Heterogeneous Integration Competence Center and the NILPhotonics Competence Center, support materials and process integration work.

The relevance of each is specific. Bonding determines whether a process can join two dissimilar materials with the precision and cleanliness a quantum device needs. Lithography determines whether a tool can pattern optical structures and control features at the required resolution, and nanoimprint offers a route to sub-wavelength structures without the cost of a projection tool. Thin-wafer processing determines how much material a process can remove while keeping the device intact, which matters for stacking.

Taken together, this is a portfolio aimed at one problem: joining and patterning dissimilar materials at high precision without degrading the properties that make them useful in the first place.

The competence centres

The two competence centres deserve a sentence each. The Heterogeneous Integration Competence Center focuses on joining dissimilar materials, the core problem for photonics and for any architecture combining more than one substrate. The NILPhotonics Competence Center focuses on nanoimprint lithography for photonic structures, where the requirement is sub-wavelength patterning over large areas rather than the finest possible feature size.

Neither is quantum-specific. That is the point: the centres existed before quantum became a market, and the quantum work applies capabilities developed for other purposes entirely, which is a more defensible position than building a dedicated facility for a market that may never arrive in the form anyone expects.

The trapped-ion case and CHAMP-ION

The trapped-ion work is the most concrete commitment in the release. EVG says it is a strategic partner of Silicon Austria Labs in support of the CHAMP-ION pilot line project, described as Europe’s first open-access pilot line for industrial-scale production of ion-trap quantum chips.

An open-access pilot line is a different kind of commitment from a customer relationship. It means equipment and process expertise go into a shared facility that multiple companies and research groups can use, which is how a manufacturing ecosystem forms around an architecture before any single company has a product. If Europe wants ion-trap quantum capability, it needs a line that can produce chips for whoever is developing them, and that is what the project describes.

The Xanadu collaboration, mentioned briefly, points the same direction in photonics: a partnership with a quantum photonic computing developer rather than a sale to one.

The market projection

The release cites McKinsey’s Quantum Technology Monitor 2026 for a projection that quantum computing could create up to 2.7 trillion dollars of economic value worldwide by 2035, with a link to the source.

That figure is worth reporting with its attribution and its limits. It projects economic value rather than revenue, it spans a decade, and it comes from a consultancy rather than from observed adoption. It is the kind of number that justifies investment in a supply chain before the demand is visible, which is precisely how it functions in a release like this one, and whether the figure serves investment or valuation is a distinction worth keeping.

Quantum Computing Scaling Is a Packaging Problem, EV Group Argues
Photo by Growtika on Unsplash

What is missing

The release describes no customer deployment beyond the Xanadu collaboration, which appears without terms. No yield data, no process specifications, no wafer sizes and no indication of what precision EVG achieves on alignment or bonding. The CHAMP-ION pilot line has no timeline, no capacity figure and no participants listed beyond EVG and Silicon Austria Labs. And the release does not say which of the processes described are in production for quantum customers today rather than in development. It also gives no indication of cycle time, which for a manufacturing process matters as much as yield: a bond that works but takes hours per wafer is a different proposition from one that takes minutes.

What to watch

Whether CHAMP-ION publishes a timeline and a participant list, because an open-access line becomes useful only once someone can book time on it. Whether EVG discloses yield or precision figures for quantum bonding, since those determine whether a process is manufacturable rather than demonstrable. And, whether the Xanadu collaboration produces a shipped device. Whether a quantum-specific tool emerges from this portfolio, since at present the offering is general-purpose equipment applied to quantum problems rather than equipment designed for them. And whether the industry converges on one architecture faster than expected, because a supplier serving six is hedged and a supplier serving one is exposed, and hedging carries its own cost in focus and in depth of engagement.