Who Really Wins If Trapped Ions Win

In September 2025, IonQ paid about $1.075 billion for Oxford Ionics, a University of Oxford spin-out with no large machine to sell. The money went to a method for getting rid of lasers.

That deserves a moment of explanation, because trapped-ion quantum computing is a laser-and-atom business at its core. Every commercial trapped-ion system running today uses precisely tuned laser beams to cool, initialise, manipulate and read out individual atoms suspended in electromagnetic fields. The lasers are what make the machines work. They are also what makes them expensive, alignment-sensitive, and slow to build in quantity.

Oxford Ionics, founded by Chris Ballance and Tom Harty, developed an approach it calls electronic qubit control, or EQC: quantum logic gates driven by microwave signals delivered through electrodes patterned into a semiconductor chip, instead of by laser beams aimed at individual atoms. The chips are fabricated by Infineon. If the approach holds at scale, it converts the central engineering problem of this modality from an optics problem into a chip problem.

That is a supply chain claim before it is a physics claim, and it is why the question in the title has no settled answer. Value in trapped-ion computing concentrates at the control layer, and the control layer is currently contested between two industrial bases that share almost nothing.

We map the stack here layer by layer, and at each layer we ask the same three questions: who supplies it, who could be squeezed by it, and who books the revenue if trapped ions scale. A longer technical treatment of the modality itself sits on PostQuantum.com. Nothing below is investment advice.

What Sits Inside a Trapped-Ion Machine

Take an atom of ytterbium, barium, calcium or strontium and strip away one electron. What remains is an ion, a charged particle that can be held in place by electromagnetic fields. Suspend a row of them in a vacuum chamber, cool them with laser light until they barely move, and encode a quantum bit in two of the ion’s electronic energy levels. Manipulate that information with laser pulses or microwave fields, then read the answer by collecting the photons the ions emit.

Every ion of a given species is identical to every other one. No fabricated qubit can say the same, and that natural uniformity is the source of the modality’s headline advantage: the highest reported gate fidelities in the field, and coherence times measured in seconds rather than microseconds. Gate fidelity is simply the share of operations that complete without introducing an error, and at these numbers the errors are counted in parts per ten thousand.

A working system needs six things:

• A trap chip. A microfabricated structure, usually on silicon or fused silica, whose electrodes generate the fields that confine the ions and move them around.
• A laser system. Several lasers at different wavelengths for photoionisation, cooling, state preparation, gate operations and readout.
• An ultra-high vacuum chamber. Pumped down to roughly fourteen orders of magnitude below atmospheric pressure, so that a stray gas molecule hits a qubit only once every several minutes.
• Optical delivery and collection. Precision optics to route beams onto ions a few micrometres apart, acousto-optic modulators to switch those beams on and off in microseconds, and high-collection-efficiency imaging to catch the fluorescence coming back.
• Control electronics. Radio-frequency drive for the trap, dozens to hundreds of low-noise voltage channels for the DC electrodes, and field-programmable gate arrays sequencing everything in real time.
• A classical layer. Calibration, compilation, and error-correction decoding.

There is no dilution refrigerator, the machine that takes superconducting processors down to roughly ten millikelvin and defines that modality’s logistics. Quantinuum’s Helios cools its trap to around 15 K, which a closed-cycle cryocooler can deliver, and most other trapped-ion systems run at or near room temperature. The helium-3 dependency disappears. In its place comes a dependency on precision photonics, ultra-high vacuum engineering, and the atomic physics expertise to hold it all in tune for hours at a time.

The Control Layer Is the Contested Layer

In a laser-controlled machine, each additional qubit tends to demand another addressed optical channel, another alignment, and another stabilisation loop. That is the trapped-ion equivalent of the wiring bottleneck that constrains superconducting processors inside a single cryostat. Three approaches are now competing to break it, and each one hands the value to a different set of suppliers.

Three routes out of the beam problem

Keep the light, shrink the optics. Photonic integrated circuits move laser routing off the optical table and onto the chip, replacing mirrors and fibre with waveguides. Groups at Sandia National Laboratories and MIT Lincoln Laboratory have demonstrated ion traps with integrated photonics, and defence research funding in the United States is pushing toward foundry-scale fabrication of these devices at ultraviolet and visible wavelengths, which is far harder than the telecom-band photonics the industry already knows how to make. If this route wins, laser and photonics companies capture the growth.

Replace the gate lasers with microwaves. Oxford Ionics’ EQC, now the centre of IonQ’s roadmap, drives the gates from electrodes built into the chip. Lasers do not disappear, but the most alignment-sensitive and least scalable component does. If this route wins, semiconductor fabrication captures the growth.

Use magnetic gradients. eleQtron, a spin-out from the University of Siegen, uses radio-frequency fields together with a strong magnetic field gradient to couple ions, an approach it calls MAGIC, for magnetic gradient induced coupling. Different physics from EQC, same commercial consequence: the hard part moves from the optics bench to the electronics.

Our reading is that the honest framing isn’t lasers versus no lasers. Cooling, photoionisation, state preparation and readout all still need light, in every architecture on the market. What is genuinely in play is the number of individually addressed beams per qubit, and whether that number can be driven toward zero.

Lasers, and the barium decision

The clearest illustration of supply chain thinking inside this modality is a choice of atom.

Ytterbium ions need laser light at 369 nm for cooling and detection, 399 nm for photoionisation, 935 nm for repumping, and 355 nm for the Raman transitions that execute gates. Two of those are ultraviolet. UV lasers are expensive, limited in power, and unkind to optical coatings and to the trap surface itself. Barium’s relevant transitions sit at 493 nm, 614 nm, 650 nm and 1762 nm, in spectral regions where mature, high-power, well-characterised laser technology already exists, much of it inherited from telecommunications and industrial optics.

Quantinuum moved to barium for Helios. That was a procurement and manufacturability decision as much as a physics one, and it points at where cost curves in this modality actually come from.

The dominant supplier of frequency-stabilised laser systems for ion trapping is TOPTICA Photonics of Munich, whose tunable diode lasers and rack-integrated laser stacks appear in most serious trapped-ion laboratories worldwide. Menlo Systems, also near Munich, supplies the optical frequency combs used to stabilise those lasers against an absolute reference. MOGLabs in Melbourne serves the same market. Around them sit the broader photonics suppliers, Coherent, Thorlabs and NKT among them, plus the specialist makers of acousto-optic and electro-optic modulators.

None of this is concentrated the way cryogenics is. The industrial laser market is enormous, and quantum computing is a rounding error inside it. But the slice these machines need is narrow: few companies make ultra-stable, narrow-linewidth systems at this quality, and none of them makes them at the volumes a scaled industry would require.

The Trap Chip and the Infineon Question

Early ion traps were machined metal electrodes assembled by hand in university laboratories. Commercial systems now use microfabricated surface traps: planar electrode structures patterned on silicon, glass or fused silica with techniques taken directly from semiconductor manufacturing. A modern trap can carry hundreds of independently driven DC electrodes for moving ions around, radio-frequency electrodes for confinement, through-substrate vias, and in research devices, integrated waveguides and photon detectors.

One company has quietly become the default foundry for these chips. Infineon Technologies has run a trapped-ion fabrication programme at its Villach site in Austria since around 2017, processing wafers across a range of substrates and diameters. It fabricates for Oxford Ionics and therefore for IonQ. It supplies eleQtron. It works with ZuriQ, an ETH Zurich spin-out pursuing micro-Penning trap designs. It fabricates for German and Austrian academic programmes. And in November 2024 it announced a partnership with Quantinuum on next-generation traps, which is the most telling of the set, since Quantinuum has in-house fabrication through Honeywell and did not need an outside foundry to make its current devices.

That combination puts Infineon in roughly the position Bluefors occupies in superconducting quantum computing: one supplier woven into nearly every serious roadmap in the modality. We would not push the parallel too far. Bluefors is a quantum company whose fortunes rise and fall with the sector. Infineon’s quantum unit is small against its automotive and power semiconductor businesses, and that cuts both ways. Quantum demand cannot starve Infineon of capacity, and quantum demand also cannot compel Infineon’s attention if a power electronics cycle turns.

Two other sources of fabrication matter. Honeywell makes Quantinuum’s traps in-house, which is a vertical integration advantage no competitor can copy quickly. Sandia National Laboratories operates one of the most advanced trap fabrication lines anywhere, and Sandia describes its Enchilada Trap as holding up to 200 ions, but Sandia serves United States government programmes rather than the open market. Europe’s answer is CHAMP-ION, a pilot line coordinated by Silicon Austria Labs, which applies to trap chips the same industrial policy logic the EU has applied to semiconductors generally.

For anyone building a position, the trap foundry is the layer with genuine power over industry scaling. It is also the layer where the financial exposure is thinnest, since trap chips will remain immaterial to Infineon’s earnings for years. The exposure is strategic rather than financial, and those are different investments.

Where There Is No Bottleneck

Two layers of this stack are strategically boring, and that is worth stating plainly, because absence of a bottleneck is genuine information.

Vacuum. Ultra-high vacuum technology has served particle physics, semiconductor manufacturing and surface science for decades. Pfeiffer Vacuum, Edwards, Agilent, Kurt J. Lesker, VACOM and SAES supply chambers, pumps, getters and viewports into a competitive market with no chokepoint. What is hard about vacuum in this application isn’t procurement, it’s assembly. Reaching and holding that pressure requires meticulous cleanliness, multi-day bakeouts, and materials chosen for very low outgassing. That work is labour-intensive and resists automation, which is one reason Alpine Quantum Technologies’ achievement of packing a complete system into two standard 19-inch racks at under 2 kW counts as a vacuum engineering result and not only an optical one.

Ion source materials. Ytterbium, barium, calcium and strontium are ordinary laboratory chemicals in the quantities these machines consume. There is no exotic materials bottleneck at the qubit layer at all. Compare the superconducting side, which depends on helium-3 produced as a decay product in nuclear weapons programmes and on niobium, whose supply is overwhelmingly concentrated in one country.

The consequence for investors is unsentimental. No scarcity means no pricing power. These layers are a structural advantage for the modality and a poor place to look for returns, and we would treat any pitch built on quantum demand for vacuum components with corresponding scepticism.

Two Platforms, Two Supply Chain Strategies

At the system level, trapped-ion computing is more concentrated than superconducting. Two companies dominate commercial supply, and they have made opposite bets about how to build a supply chain.

Quantinuum, built inside

Quantinuum runs the full stack. Honeywell fabricates the traps, the laser architecture and control system are proprietary, and the software layer, including the TKET compiler and the Quantum Origin key generation product, is developed in-house. Helios, launched in November 2025, runs 98 barium-137 qubits in a QCCD architecture, short for quantum charge-coupled device, in which ions are physically shuttled between separate operating zones of one chip rather than sitting in a single fixed chain.

Quantinuum reports 48 error-corrected logical qubits from those 98 physical ones in its Helios launch materials, a two-to-one encoding ratio. This number needs careful handling, and it is the kind of number we spend real time on with candidates.

A physical qubit is a single ion. A logical qubit is an abstraction built from several physical qubits plus a code that detects and corrects errors, and the number of physical qubits each logical one consumes depends entirely on the code distance, which is roughly how many simultaneous errors the code can survive. Resource estimates for a machine running Shor’s algorithm against RSA-2048 assume deep codes with overheads in the hundreds or thousands to one. A two-to-one ratio comes from shallow codes, and it is a real and impressive result about physical fidelity: when your raw error rate is low enough, cheap encodings become useful. It is not the ratio that will apply to a cryptographically relevant machine. Announced roadmaps, Sol and Apollo among them, are targets rather than results, and should be read as such.

Honeywell holds a majority of Quantinuum, so public-market exposure to the company today runs through Honeywell. Quantinuum announced in 2025 that it had raised $600 million at a $10 billion pre-money valuation, with NVIDIA’s venture arm among the investors.

IonQ, built by acquisition

IonQ took the other route. Its announced transactions through 2025 include Oxford Ionics for control technology and a fabrication path, Lightsynq for quantum memory and photonic interconnect work, Qubitekk for quantum networking, Vector Atomic for precision timing and inertial sensing, and a majority position in the Swiss quantum communications firm ID Quantique.

Its published roadmap escalates from hundreds of physical qubits to tens of thousands on interconnected chips within a few years, and to two million physical qubits by 2030. Every step in that sequence depends on something not yet demonstrated at commercial scale: trap chips with thousands of electrodes manufactured reliably and affordably, photonic interconnects that distribute entanglement fast enough to be useful, and a classical control architecture capable of orchestrating all of it.

The two companies are placing their risk in different accounts. Quantinuum’s risk is being out-scaled by a competitor willing to move faster than its own manufacturing discipline allows. IonQ’s risk is integration, because five acquired engineering cultures have to converge into one product line while the roadmap keeps its promised slope. Neither is the safe option.

Europe Builds the Parts, America Owns the Platforms

Line up the layers and something unusual appears. Infineon fabricates the traps in Austria. TOPTICA supplies the lasers from Munich, Menlo the frequency combs. Pfeiffer and VACOM supply vacuum hardware from Germany. Alpine Quantum Technologies in Innsbruck, founded by Rainer Blatt and Thomas Monz, builds complete calcium-ion systems that fit in server racks. eleQtron builds microwave-controlled systems in Siegen. Universal Quantum in Brighton, out of Winfried Hensinger’s group at Sussex, pursues a modular architecture that moves ions physically between chips. ZuriQ works out of Zurich. The research lineage running through Innsbruck, Oxford, Sussex, Mainz and ETH is where most of the field’s founding physicists trained.

Public money reinforces all of it. Germany’s ATIQ programme, a federally funded consortium building an applications-oriented trapped-ion quantum computer, is building a national demonstrator with TOPTICA, eleQtron and Infineon among the partners, and the Federal Ministry of Education and Research funds it at roughly €45 million. CHAMP-ION is building the fabrication pilot line. EuroHPC procurement has put AQT machines into European supercomputing centres, including a system in Poland inaugurated in 2025 and a delivery to the Leibniz Supercomputing Centre in Munich. Germany’s Cyberagentur has funded work on transportable ion-trap machines, including at Universal Quantum in the United Kingdom.

This is the most coherent regional supply chain in quantum computing, and it exists because trapped ions play to what Europe already has: precision photonics, semiconductor fabrication, and atomic physics. Europe does not build dilution refrigerators at scale. It does build these.

There is also a deployment argument that matters more to sovereignty than component share does. A machine that runs at room temperature, fits in two 19-inch racks and draws under 2 kW can be installed inside a national computing centre and operated by that centre’s own staff, with data never leaving the jurisdiction. A national laboratory can own a trapped-ion computer. Renting cloud access to a machine on another continent is a different proposition, and for classified or regulated workloads it is often not a proposition at all.

Then there is the other side of the ledger. Oxford Ionics was a British spin-out running on Austrian silicon, and it is now a division of a New York-listed company that has also announced a majority position in Swiss-based ID Quantique. The pattern is familiar from semiconductors and from AI: the components and the physicists are European, the platform and the customer relationship increasingly are not. If you are modelling sovereignty exposure rather than equity exposure, the question to ask is who holds the roadmap the chip is being fabricated for, not who fabricates it.

The Layer That Does Not Exist Yet

Every trapped-ion roadmap that reaches thousands of qubits depends on a technology that has not been built at commercial scale.

A single chain of ions becomes unmanageable somewhere past roughly 50 to 100 ions. QCCD shuttling extends the ceiling by moving ions between zones of one chip, but chip complexity climbs steeply. The route past that ceiling is modularity: link separate trap modules so that they behave as one computer. The standard method is a photonic interconnect. An ion in one module emits a photon, the photon travels down an optical fibre, and a joint measurement establishes entanglement between ions sitting in two different modules, which lets quantum information move between them.

Making this work drags a cryogenic dependency back into a room-temperature modality. Detecting single photons at the required efficiency generally means superconducting nanowire single-photon detectors, SNSPDs, which operate at a few kelvin and need their own cryocoolers. Single Quantum in the Netherlands, ID Quantique in Switzerland and Photon Spot in the United States supply them.

It also creates demand for quantum memory, a device that holds a qubit steady while remote entanglement attempts repeat until one succeeds. That was the logic behind IonQ’s purchase of Lightsynq, and it is among the least mature components anywhere in the stack.

Universal Quantum’s alternative avoids photons entirely by transporting the ions themselves between modules along electric-field highways. It trades an unsolved photonics problem for an unsolved ion-transport problem, which is a legitimate bet rather than a shortcut.

Demonstrated entanglement rates between remote ion modules remain orders of magnitude below what a machine built from thousands of modules would need. This is the single largest uncertainty in the modality, and it is where the gap between roadmap slides and laboratory results is widest.

What Would Derail This

Interconnects stall. If remote entanglement rates stay where they are and single-chip qubit counts plateau in the hundreds, trapped ions remain a high-fidelity, small-machine business. That is a real market, particularly for chemistry and for error-correction research, but it is not the market the current valuations describe.

Laser integration takes longer than expected. If optical tables cannot be compressed into rack-mounted or chip-scale packages on schedule, cost and footprint keep these systems in laboratories and cloud queues rather than in customer data centres, and the sovereignty argument above weakens with them.

Infineon reprioritises. Trap fabrication is a small line inside a large semiconductor company. A capacity squeeze or a strategic reshuffle would hit IonQ, eleQtron and ZuriQ simultaneously, and only Quantinuum has an in-house alternative.

Neutral atoms scale first. Neutral-atom machines from QuEra, Pasqal and Infleqtion draw on the same lasers, the same vacuum technology and much of the same optical control, while scaling through optical tweezer arrays rather than electrode structures. That comparison contains the most useful hedge in this whole analysis. The shared layers, lasers, photonics and vacuum, sell into both modalities and cannot be stranded by the outcome. The ion-specific layers, trap design and species expertise, can be.

Questions Worth Asking a Vendor

If you are evaluating a trapped-ion partner, a procurement conversation, or a pitch deck, four questions separate substance from slide.

1. Physical or logical? Ask for both counts, and then ask for the code distance behind the logical number. A two-to-one encoding and a thousand-to-one encoding describe different machines and different decades.
2. Announced or demonstrated? Every roadmap milestone falls into one bucket. Ask which results have been published, which have been measured internally, and which are plans.
3. What in this roadmap depends on interconnects? If the answer is everything past the next generation, price that accordingly.
4. Who fabricates the trap, and what is the second source? For most of this field the honest answer today is one foundry and no alternative.

Modality literacy is a procurement skill, not a physics hobby. The teams who get quantum decisions right over the next few years will be the ones who can read a vendor roadmap the way they read a supplier audit: layer by layer, dependency by dependency, with a clear view of which claims have been demonstrated and which have merely been scheduled. Building that habit, layer by layer and dependency by dependency, is what our hardware and supply chain training is for, and the current program catalogue is at quantumacademy.com/.