A quantum interconnect is a device that lets two quantum computers exchange information with each other, the same way a network cable lets two ordinary computers talk. It sounds simple. It is one of the hardest unsolved problems in quantum technology.
Here is the short version. Quantum computers today are built from superconducting circuits that only work near absolute zero, inside a refrigerator called a cryostat. To send information over long distances, you need light traveling through fiber optic cable, the same fiber that carries the internet. The catch: a single particle of that light carries roughly 40,000 times more energy than a single particle of the microwave signal a qubit uses. Turning one into the other, without losing the fragile quantum information along the way, is the whole challenge.
A quantum interconnect is a device that lets separate quantum computers exchange information without breaking the fragile quantum state that makes them useful, and nobody has built a working one yet. The hard part is converting between the microwave signals a qubit uses and the light signals a fiber optic cable carries, without losing too much signal, adding too much noise, or needing extreme cooling to do it. Researchers are pursuing five distinct physical approaches to that conversion, and four of them each solve part of the problem while running into a different limit of their own. This page makes the case for the fifth: a hybrid light and matter particle called an exciton-polariton, sometimes described as a quantum polariton interconnect, which sidesteps that tradeoff because the conversion is built into the particle itself, rather than routed through a separate mechanism.
The end goal is what you might call a fiber networked quantum data center, the infrastructure behind modular quantum computing. Picture several quantum computers, each in its own cryostat, connected by ordinary fiber optic cable, working together closely enough to correct each other's errors, the way separate machines in a data center already do today. Once that works for one type of quantum computer, the same idea can connect other types too, including trapped-ion systems and other quantum hardware that doesn't use superconducting qubits at all.
Why not just run a cable of microwave wiring between two cryostats instead of solving this optical conversion problem? That approach works, but only across a distance of about a meter, and the cooling cost rises quickly as the distance grows. Fiber optic cable is the only medium already proven to carry signals long distances without much loss, at room temperature. The problem is that a quantum computer does not speak light natively. Something has to translate, a process researchers call cryogenic microwave-to-optical transduction. A national roadmap from the Department of Energy's Q-NEXT center, based at Argonne National Laboratory, puts it directly: "microwave-to-optical telecom wavelength transduction is important given the relevance of superconducting qubit-based computing."[1]
Researchers judge every proposed interconnect on nine measurements, drawn from the field's own framework for characterizing transducers.[2] The hard part is not achieving any single one. It is achieving all critical requirements together, because improving one measurement usually makes another one worse.
Researchers have studied this conversion problem for about twenty years, and made real progress, including a detailed 2026 review from scientists at Fujitsu Research.[3][4] A community roadmap sponsored by the National Science Foundation and published in PRX Quantum put it plainly: "the quantum interconnect bottleneck is imminent, and is emerging as a grand challenge" for the whole field.[5] Investor commentary frames the stakes just as directly: "quantum interconnects are critical devices that allow the transfer of quantum information," writes venture investor Russ Fein.[6] Yet no one has turned that progress into a working, ready to install interconnect. The missing piece is not a single scientific breakthrough. It is that solving this problem end to end takes far more disciplines than any one team usually covers.
Building one working interconnect means starting from the basic physics of the device, then working through materials science, microwave engineering, and photonics, then packaging the whole thing to survive inside a cryostat, then proving it works between two real quantum computers. Each of those steps is its own specialty, usually studied by a different lab, with its own training, its own equipment, and its own way of measuring success. A materials scientist who can grow the right crystal rarely also builds microwave circuits, and a photonics engineer rarely also runs a dilution refrigerator. A device that works beautifully on an optical bench can still fail the moment it has to survive real cryogenic packaging, and a device that survives packaging still has to be proven between two actual quantum computers, not just in isolation. Very few teams anywhere hold every one of those skills under one roof at once.
Government and university funding is set up to fund each piece separately, a better material here, a better resonator there, rather than one team owning the whole chain. Grants are typically sized and timed around a single component or a single publishable result, which rewards the next incremental improvement far more reliably than it rewards the harder, slower work of integrating many components into one finished device. That mismatch compounds over time: each grant cycle produces another improved piece, but nobody's funding is ever structured to pay for the unglamorous work of making all those improved pieces fit together. Companies building quantum computers, meanwhile, are focused on their own processors, not on shared, open infrastructure that would let different quantum computers connect to each other, since a proprietary processor is easier to justify investing in than infrastructure the whole field would benefit from equally. The result is a lot of excellent components and very little that connects them.
Even when one lab's piece of the puzzle succeeds, on its own that success does not become a working interconnect. A published result usually proves one stage works in one lab's specific setup, not that it survives being handed off to the next stage, run by a different team with different equipment and different assumptions. Turning individual results into one validated, installable device takes a team with the mandate and the mix of skills to own every stage at once, from the first physics result through packaging to a working link between two real machines, and no existing lab, company, or funding program currently holds that mandate. Universities reward individual investigators and individual papers, national laboratories are not typically structured to carry a device through commercialization, and companies have the mandate to ship a product but usually only for their own processor. Large industry players have recently started publicly soliciting outside research into exactly this gap, itself a sign of how unmet the need still is.[7] Industry investment has kept accelerating since: in 2026 Cisco unveiled a prototype Universal Quantum Switch designed to route and convert signals between different quantum computing modalities at the network layer, with Cisco Research Fellow Ramana Kompella noting that "quantum networking has many practical and commercial use cases in the classical world today."[8] That switch sits one layer above the microwave-to-optical, qubit-level transduction problem this page focuses on, but it reflects the same trend: the interconnect gap is now serious enough that major infrastructure companies are building dedicated hardware for it, not just asking outside researchers to.
The result is a technology that the entire field needs, but that no single lab, company, or funding program currently owns from the physics all the way to a finished, working device.
Researchers around the world are trying several genuinely different physical approaches to this conversion problem. Each one has produced real, published results. Each one also runs into a limit that comes from the approach itself, not just from needing more engineering time. The diagram in each card shows, in simple form, how the approach works and where it runs into trouble.
A microscopic drum or beam sits between the microwave circuit and the optical cavity. It absorbs the microwave signal as physical motion, then that motion nudges the light. This method was the first to ever convert a real qubit's signal into light, and it still holds the record for the highest conversion efficiency of any method.[9] Related devices have since pushed the same mechanical approach further: a fully integrated microwave-to-optics chip,[10] a transducer that reads out a superconducting qubit optically,[11] and a silicon-based version reaching deeper into the quantum-enabled regime.[12]
This approach has already reached the commercial market. In March 2026, the Delft-based company QphoX, a TU Delft spin-off, launched a microwave-to-optical quantum transducer built on this same optomechanical foundation as a commercial product. IBM became its first partner, testing the device on IBM's own Quantum Networking Unit hardware as part of a broader push toward distributed, modular quantum computing. Jerry Chow, IBM's CTO of Quantum-Centric Supercomputing, has framed the partnership as a way to explore whether transducer technology like this could help IBM scale its quantum computers beyond its existing roadmap, toward genuinely distributed networks of machines.[13]
A microwave field directly reshapes light passing through a special crystal, with no moving parts and no mechanical middle step at all. The conversion happens through the material's optical properties changing under the microwave field.[14] A pair of these devices has already linked two separate superconducting circuits, each in its own dilution refrigerator, over a kilometer of fiber, confirming the basic two-transducer architecture works at real distance, though still at classical signal levels rather than the single-photon regime a working interconnect needs.[15] A newer variant pushes the same crystal-based approach up to millimeter-wave frequencies, trading some efficiency for the chance to run at a warmer, easier to cool stage of the system.[16]
Instead of one particle doing the conversion, a large crowd of spins or atoms does it together. Working as a group lets an otherwise weak signal add up into something usable, a bit like a stadium crowd being louder together than any one person shouting alone.[17] Recent work has started closing the chip-integration gap for rare-earth systems specifically, embedding erbium in foundry-fabricated photonics while preserving long optical coherence.[18]
This broader family routes the conversion through other collective behaviors in materials, like spin waves in magnetic crystals, or sound waves paired with light. One recent example coupled spin waves to excitons in a layered antiferromagnet and found a broad, roughly 300 megahertz operating window.[19] Another created a hybrid quasiparticle, part photon, part phonon, part exciton, showing coherent two-way conversion through sound rather than an electrical drive, though without a reported efficiency or noise figure.[20] These experiments are useful because they prove a wider idea: mixing light and matter tightly, inside a well built cavity, is a real route to efficient conversion.
Every method above routes the conversion through some go between: motion, a weak material effect, or a crowd of atoms. This approach skips the go between entirely, and the same device runs the conversion in both directions. Read the full explanation below.
Look at the pattern across the first four approaches. Each one buys its strength by paying a cost somewhere else in the system, in speed, in efficiency, in temperature, or in added heat. That trade off is exactly why no single platform has yet hit all nine measurements from the earlier section at the same time. A 2025 theoretical study reached a similar conclusion independently, modeling a different platform (color-center ensembles) and finding that strong enough cavity coupling could push efficiency above 90 percent with vanishingly little added noise, the same underlying lesson this landscape keeps pointing to.[21]
A polariton forms when a photon (a particle of light) trapped inside a semiconductor optical microcavity, a very high quality optical cavity built into a semiconductor chip, mixes so strongly with an exciton (a paired electron and its missing partner, inside that same semiconductor) that the two stop being separate things. What comes out is one particle that is genuinely part light and part matter, at the same time. There is no go between step to add loss, slow things down, or leak heat, because the mixing is built into the particle itself. An incoming microwave photon interacts with the polariton through its exciton half, a direct microwave-polariton coupling, and light leaving the cavity carries that signal's amplitude and phase back out. Run the same device in reverse at the receiving end, and it turns the light back into a microwave photon able to excite a second qubit. Every step of that trip is coherent, so it is the quantum state itself, not just a copy of the signal, that survives the journey from one qubit to the fiber and on to another, making this a genuinely bidirectional quantum interconnect.
Every other method spends part of its signal driving an intermediary, a drum, a crystal, a crowd of atoms, before that intermediary can pass anything to the light. The polariton is already both light and matter mixed together, so more of the original signal survives the trip, rather than being spent along the way.
The same physical particle converts microwave to light heading out, and light back to microwave heading in. There is no separate transmit device and receive device to build, calibrate, and keep in sync. One converter handles the full round trip a real quantum network needs.
Mechanical approaches need a strong pump to drive the intermediary, and that pump energy leaks back into the cold qubit as unwanted heat. The polariton's mixing comes from the cavity and material design itself, not from an external drive, so there is far less added heat for the cryostat to fight against.
same qubit, same destination fiber, a very different number of handoffs in between
Certain engineered versions of these particles, called trion polaritons and dipolaritons,[22] carry an electric charge or a strong electric dipole. That means they respond directly to a gigahertz microwave field, the same kind of signal a superconducting qubit produces. No conversion step is needed to make the particle sensitive to the qubit's signal in the first place, and the strength of that response can even be tuned electrically.[23] On the superconducting side of that interface, resonator fabrication techniques continue to mature in parallel, including new patterning methods for high quality-factor niobium resonators.[24]
Under the right conditions, a large number of these particles settle into a single, shared, coherent state, a process called polariton condensation, all acting together rather than independently, in much the same way photons inside a laser lock into one synchronized wave rather than many independent flashes. That collective behavior naturally boosts the strength of the conversion, since many particles responding in step add up rather than average out, and it keeps the signal clean and stable rather than smearing it out over time. Laboratory measurements of these condensates have already recorded emission linewidths narrow enough, and coherence times long enough, to stay well resolved from microwave frequencies.[25][26]
The main material used to build these particles, a gallium arsenide semiconductor structure, is already grown using standard, industrial scale manufacturing methods, and can be laid out with multiple channels on a single chip, the same wavelength-division multiplexing approach that lets one fiber carry many independent signals at once. Its natural light output sits at a wavelength with an already proven path to the 1550 nanometer band used across the entire fiber optic network.[27] The underlying physics has also been shown to work in more than one type of material, so there is a real fallback option, not just one narrow path.
This is not a speculative or unproven idea. The basic phenomenon, a shared quantum state formed from many polaritons acting as one, was first demonstrated in a semiconductor microcavity in 2006, and has since been reproduced and extended by independent labs across multiple material systems.[28] The particle itself is mature, well studied science, even though using it as a quantum interconnect is new.
The version described here operates at 1 to 4 Kelvin, but that is not a hard ceiling set by the physics. The same gallium arsenide material system used here has already been shown to support this kind of condensation at room temperature, real evidence that there is headroom to push operation warmer as the engineering matures, not just a hope.[29]
Polaritons are not confined to delicate, laser-pumped lab setups. Electrically driven polariton lasers, devices switched on with an ordinary electrical current instead of a separate pump laser, have already been built and operated, including at room temperature.[30] The basic engineering of driving a polariton device with electricity is already solved territory, not something still to be invented.
This is not only a theoretical idea. Cryogenic experiments on an atomically thin semiconductor have already shown that applying a gigahertz microwave field directly and measurably changes how the material's excitons emit light. That is a direct, hands on confirmation that microwave fields and excitons interact at exactly the frequencies a real interconnect would need, using a working cold laboratory setup that combines microwave delivery with light measurement in the same device. Figure: microwave induced change in exciton light emission, measured in a cryogenic setup on an atomically thin semiconductor. This confirms the interaction exists. It is an early measurement, not yet a calibrated conversion efficiency.
None of this is a small tweak to an existing converter. It is a genuinely different physical mechanism for closing the same gap. Turning that physics into a finished, deployable interconnect still takes the same coordinated work in materials, microwave engineering, photonics, cryogenics, and packaging that every method in this landscape ultimately has to go through. Worth saying plainly: unlike the four methods above, this page does not yet cite a calibrated end-to-end efficiency or noise figure for the polariton platform itself, since that work is still underway. The case made here is architectural, why the mechanism avoids a limitation the other four share, not a claim that it has already been measured to outperform them.
The first practical use for a link like this is heralded entanglement between two qubits, a protocol built to succeed even before conversion efficiency is perfect, since it only keeps the runs that work and discards the rest. As efficiency climbs closer to the point where a single attempt reliably succeeds, the same hardware opens the door to deterministic, on-demand state transfer between qubits, no discarding required. Because the underlying mechanism is a general recipe for coupling microwave fields to light rather than something built around one specific qubit design, the same approach extends in principle to links with atomic systems, solid-state defect qubits, and other photonic quantum hardware, not only superconducting processors.
The core exciton-polariton interconnect approach described above is covered by issued U.S. patents, Kolmakov, G.V. et al., U.S. Patents 11,874,581 B2 (2024) and 12,481,200 B2 (2025), exclusively licensed to QuantumCue for broad industry adoption.[31]
This is not a niche problem. The United States has formally named quantum networking a national research priority, and the government's own most recent assessment names the exact device category this page is about.
The National Quantum Initiative Act, signed into law in 2018, made quantum information science a coordinated federal priority across agencies. In 2020, the Department of Energy (DOE) and its national laboratories published a formal blueprint for a national quantum internet, naming the research needed to connect quantum systems over real infrastructure as one of the country's foundational science objectives.[32]
In September 2024, the National Quantum Initiative Advisory Committee, which reports its findings to the President and Congress, published a dedicated assessment of quantum networking. It concluded that quantum networking capabilities will affect U.S. economic prosperity and national security, though the full scale of that effect will only become clear through continued research, and it named transducers, devices that convert a quantum signal from one physical form into another, specifically among the components where progress is still needed.[33] A microwave-to-optical quantum interconnect, the subject of this page, is exactly that kind of device.
That priority accelerated through 2026. In June, the President signed an executive order on quantum information science, and the Department of Energy followed with a new Quantum Genesis initiative aimed at a fault-tolerant quantum computer by 2028, backed by a proposed 1.2 billion dollar budget request for a new federal office overseeing AI and quantum research.[34] Earlier that year, the Department's broader Genesis Mission had already opened a dedicated funding track for AI-assisted work on quantum networking and multi-node processor scaling.[35] The White House's own July 2026 report on the initiative confirms quantum information science as one of at least twenty national science and technology challenge areas the Genesis Mission directs the Department of Energy to pursue.[36]
The National Science Foundation (NSF) went further, announcing a 1.5 billion dollar, decade-long initiative called NSF X-Labs, built around a small number of topics chosen for their national importance. One of the first two is titled, in the government's own language, quantum systems interconnects and integrated photonics, aimed at developing the components needed to transfer quantum information and connect different kinds of quantum hardware together, part of what the program itself describes as a "roadmap for the integration of second-generation quantum systems."[37] That is the exact technology category this page describes.
The Department of Commerce added over two billion dollars in proposed CHIPS Act incentives to nine quantum companies that same month, including funding earmarked for superconducting wafer fabrication and quantum foundry capacity, the manufacturing side of the same interconnect stack.[38]
The government has also moved to lower the cost of getting a device like this built in the first place. The National Quantum and Nanotechnology Infrastructure program gives startups shared cleanroom access to more than 2,000 fabrication tools for prototyping semiconductor, photonic, and quantum devices, often for a few hundred dollars an hour, without any one company having to build its own fabrication line.[39] That is precisely the kind of shared, cross-disciplinary infrastructure the earlier section on funding fragmentation argues the field is missing.
Defense research agencies are threaded through this landscape too. The Department of Defense, now also using the secondary title Department of War (DOW) following a 2025 executive order,[40] funds quantum-relevant work through components including the Air Force Research Laboratory (AFRL) and the Defense Advanced Research Projects Agency (DARPA), alongside the Army Research Office funding that has supported some of the underlying microwave-exciton experimental work referenced on this page.
The claims on this page are drawn from published, peer reviewed or preprint research. Each method above links to its primary source, listed again here for convenience. This list covers the key results discussed and is not an exhaustive bibliography of the field.