💻 Quantum

The Quantum Internet's Most Expensive Component Just Got 1,000x Easier to Make

NIST scaled superconducting photon detectors from 100 nanometers to 0.1 millimeters, dropped the noise floor ten billion times, and opened the door to optical lithography. The per-detector cost implications are staggering.

A superconducting photon detector chip glowing faintly blue in a cryogenic chamber, with a wide flat wire strip visible under magnification against a dark metallic substrate
Dr. Iris Blackwell · Quantum · August 26, 2026

Ten orders of magnitude.

A team at the National Institute of Standards and Technology did something the superconducting detector community considered impossible: they built a single-photon detector with a wire 1,000 times wider than standard devices, and it got quieter. Not by a factor of two, or a thousand, but by ten billion. Their result, published in Optica on August 19, doesn't just set a performance record. It breaks the manufacturing cost curve for sensors that sit at the heart of quantum computing, quantum networking, secure communications, and deep-tissue medical imaging.

Because the fabrication bottleneck that has kept these detectors expensive for two decades is about to become irrelevant.

The nanowire trap

Superconducting nanowire single-photon detectors, known as SNSPDs, are the gold standard for catching individual photons. A hair-thin wire of superconducting material carries current with zero resistance at cryogenic temperatures, and when a single photon strikes the wire it creates a tiny hot spot that momentarily breaks the superconductivity, generating a detectable voltage pulse. NIST's own devices have reached 98% detection efficiency, catching nearly every photon that arrives.

But there is a problem buried in the name: nanowire. For decades, physicists believed these wires had to be roughly 100 nanometers wide, about 1/1,000th the width of a human hair, because at that scale a single photon's energy is enough to disrupt the supercurrent across the full width of the wire, producing a clean signal. Make the wire wider and the photon's splash gets lost in the current, a pebble tossed into the Mississippi.

Nanoscale fabrication means electron-beam lithography, and e-beam systems write patterns by scanning a focused beam of electrons across a surface one point at a time, like drawing with a very expensive pen. Resolution is stunning at below 10 nanometers, but throughput is abysmal at roughly 0.01 cm² per hour. Optical lithography, the workhorse of the semiconductor industry, processes wafers at rates exceeding 700 cm² across an entire wafer in parallel, and the throughput gap between the two is not a rounding error but a factor of roughly ten million.

That throughput gap explains why a commercial SNSPD system costs $80,000 to $350,000 per detection channel, according to market data from DataIntelo's 2025 market analysis. ID Quantique, Single Quantum, and Photon Spot dominate a global market worth approximately $40 million in 2026, constrained not by demand but by how hard these things are to manufacture.

Rails, not nanowires

Kristen Parzuchowski, a postdoctoral researcher at NIST, and co-lead Eli Mueller solved the width problem by adding what they call "rails." Two current-carrying superconducting strips made of niobium run parallel to the main detector wire on either side, generating a magnetic field that redistributes current flowing through the central wire, pushing it away from the edges and toward the center. Edge current crowding, the phenomenon that has limited SNSPD performance since the technology's invention in 2001, is gone.

With edges tamed, Parzuchowski and Mueller's team scaled the detector wire from nanometers to 100 micrometers (0.1 millimeters) and demonstrated single-photon detection at 1,550 nanometers, the standard telecom wavelength. A 20-micrometer-wide device achieved near-unity internal detection efficiency at 4 micrometers, a wavelength deep in the mid-infrared and a factor of 20 wider than the best prior result for that spectral range.

But here is what stunned even the researchers: rails recovered detectors that would otherwise have been scrap. One 10-micrometer device had such poor edge quality that it showed zero photosensitivity at 1,550 nm with rails off, yet turning rails on produced near-unity detection efficiency. "It was very shocking," Parzuchowski recalled, "we were calling in some other people in our group into the room, telling them, 'Look at this, this is crazy!'"

The fabrication math nobody has run

Here is the calculation that matters, and it is one nobody has published. A conventional SNSPD uses a 100-nanometer wire meandered back and forth to fill a detection area roughly matching the 9-micrometer mode field diameter of a standard single-mode fiber. That meander pattern requires electron-beam lithography because no other technique can reliably write 100-nanometer features with the edge quality these detectors demand.

At 100 micrometers wide, a single straight strip covers the fiber mode diameter with no meander at all. No meander means no bends, no corners, no edge artifacts. And 100-micrometer features are well within the capability of contact photolithography, a technique so mature and inexpensive that a functional system costs under $100,000 and university cleanrooms routinely purchase them used for $10,000 to $30,000. Compare that to e-beam: $10 million to $30 million per system.

Capital cost is not even the real story; throughput is.

ParameterE-Beam (Nanowire SNSPD)Contact Photolithography (Wide SNSPD)
Feature size100 nm100 µm
Write methodSerial (point by point)Parallel (full-field exposure)
Throughput~0.01 cm²/hourWafer-scale (seconds per exposure)
System cost$10M–$30M$10K–$100K
Meander requiredYes (hundreds of bends)No (single strip)
Pattern complexity~2,000+ parallel passes3 elements (wire + 2 rails)

A conservative estimate: if fabrication accounts for 30% to 50% of SNSPD system cost (the remainder being cryogenic cooling, readout electronics, and fiber coupling), moving from e-beam to contact photolithography could reduce per-detector fabrication cost by one to two orders of magnitude. A $350,000 detection channel does not become $350 overnight because the cryogenics alone prevent that, but the fabrication component could plausibly drop from $100,000 to $175,000 per channel down to $1,000 to $15,000. Cross that line and quantum networking operators stop buying individual detectors for lab experiments and start ordering them by the rack for infrastructure deployment.

What ten billion times quieter means for distance

Dark counts are false signals: the detector fires when no photon arrived. In quantum key distribution, maximum secure distance between nodes is fundamentally limited by the point at which legitimate photon arrivals drop below the dark count rate, and standard fiber loses about 0.2 dB per kilometer at 1,550 nm, so signal strength decays exponentially with distance.

A ten-order-of-magnitude reduction in dark counts corresponds to 100 dB of additional noise margin. At 0.2 dB/km, that buys 500 kilometers of extra reach before signal drowns in noise. Current QKD demonstrations using conventional SNSPDs max out near 500 to 600 kilometers, so in principle this dark count reduction alone could push single-span capability past 1,000 kilometers, though channel loss, timing uncertainty, and protocol overhead impose their own ceilings well before that theoretical maximum.

Even a more conservative interpretation matters enormously: if you can span 200 km instead of 100 km between trusted repeater nodes, you halve the node count in a transcontinental quantum backbone. At an estimated $500,000 to $1 million per repeater node, eliminating five nodes from a coast-to-coast link saves $2.5 million to $5 million in infrastructure, and each eliminated node is also one fewer point of vulnerability in the security chain that the whole system exists to protect.

The yield kicker

Industry analysts at GlobalGrowthInsights and BusinessResearchInsights both identify manufacturing yield as the single largest market constraint, cited by 35% of survey respondents as the primary bottleneck, and NIST's paper offers a direct answer. Rails recovered a device with a switching-to-depairing current ratio of just 44%, a ratio that rendered it completely non-functional, and restored it to near-unity detection efficiency, turning scrap into product. Applied across a production line, that kind of yield recovery transforms economics even before accounting for the lithography shift.

Limitations

Several caveats are essential. First, NIST has not yet demonstrated system detection efficiency (as opposed to internal detection efficiency) for wide devices; reaching the 98% system efficiency of the best nanowire SNSPDs requires optical coupling optimization that has not been performed at these widths, and group leader Marty Stevens explicitly noted that more testing is needed. Second, fabrication cost estimates above assume that contact photolithography can achieve the edge quality necessary for high-performance detectors, something not independently demonstrated for production SNSPDs, though the physics at 100-micrometer feature sizes strongly favors it. Third, cryogenic infrastructure required to cool these detectors to 0.9 Kelvin remains expensive and complex regardless of detector fabrication method, accounting for a substantial portion of total system cost. Fourth, array scaling from single detectors to the hundreds or thousands of channels needed for practical quantum networks introduces integration challenges that wide wires may actually help with, since they produce higher signal amplitudes that eliminate the need for cryogenic amplifiers, but nobody has tested this at scale yet.

The strongest case against

Most serious counterargument: SNSPDs have been "on the verge of transformative cost reduction" before. In 2015, tungsten silicide devices promised cheaper material alternatives to niobium nitride. In 2019, amorphous molybdenum silicide films promised simplified deposition. Both delivered real performance improvements, yet neither meaningfully changed the per-channel price tag because the e-beam lithography bottleneck remained untouched. Skeptics will argue, correctly, that a lab demonstration of optical-lithography-compatible dimensions is not the same thing as a qualified production process, and the gap between "should work" and "ships at volume" has swallowed many promising detector architectures whole. But here is what is different this time: NIST did not just widen the wire but demonstrated a mechanism (rail-tuned current redistribution) that recovers defective devices, turning what would have been manufacturing rejects into functioning detectors. Yield improvement is the one factor that compound-scales with every other cost reduction, and if it holds in production the skeptics' objection weakens considerably.

The bottom line

If you are building quantum networking infrastructure, this paper is a signal to revisit your detector cost assumptions. A $40 million SNSPD market is supply-constrained by a fabrication method ten million times slower than its replacement, and NIST just proved the replacement works at the physics level. If you are a biomedical imaging researcher working with diffuse correlation spectroscopy or other photon-starved techniques, wider detectors with sub-microhertz dark counts mean larger collection areas and cleaner signals without the current six-figure entry price. If you are an investor evaluating quantum infrastructure plays, ask the detector companies on your radar whether they have a photolithography roadmap. ID Quantique, Single Quantum, Photon Spot, Quantum Opus, Scontel. Ask them all. The ones with a roadmap are about to get very interesting. The ones without are selling quill pens in the age of the printing press.

Sources