Xanadu Optical Switch Loss Cut to 7 mdB, Firm Says

Xanadu and HyperLight say they have reduced optical switch insertion loss to 7 millidecibels (mdB), down from a 20 mdB figure the same partners reported in 2025. In a 10 November 2026 release, the Toronto photonic quantum computing company called that change “a nearly threefold improvement” and said the result was demonstrated on HyperLight’s high-volume thin-film lithium niobate Chiplet manufacturing line, with “excellent yield and repeatability.”

That 7 mdB figure is a company-reported benchmark, not an independent laboratory measurement, and the announcement was not accompanied by a paper. Xanadu’s own post on X, from the verified @XanaduAI account on 10 November 2026 at 12:13 UTC, states the same headline result: “a new performance benchmark of 7 mdB optical switch loss using a high-volume manufacturing platform.” HyperLight has not issued a separate release that TSN could find.

What the companies say the 7 mdB result is

Optical switches route photons in a photonic quantum computer. Xanadu argues that because quantum signals cannot be conventionally amplified without destroying the state they carry, insertion loss in those switches is one of the main limits on scaling. Lower loss, the release says, preserves fidelity, reduces lost qubits and avoids the need for ever more laser power or redundant hardware.

Christian Weedbrook, Xanadu’s founder and chief executive, said the 7 mdB milestone “indicates that Xanadu’s quantum-compatible switches are ready to ramp to production.” In the same release he also said the new figure is “nearly 3x lower than our previous benchmark set just over a year ago and 10x lower than our closest competitor’s most recent published result.” Those production-ready and competitor-comparison statements are Xanadu’s claims. The 10x comparison is not independently documented in the announcement, and the official X post does not repeat it.

Mian Zhang, HyperLight’s chief executive, said the same TFLN platform that the company uses for AI data-centre and telecom work had “delivered a record-setting 7 mdB switch loss for quantum computing with Xanadu.” That, too, is a partner’s characterisation of a joint result. The release describes the work as a collaboration, not as a test by an outside metrology lab.

Xanadu, which is listed on Nasdaq and the TSX under the ticker XNDU, presents the result as a step on its path to utility-scale photonic systems. The text is explicit that this is a hardware performance milestone the company itself is announcing.

A company benchmark is not a third-party measurement

The comparison that can be checked against Xanadu’s own prior claim is the move from 20 mdB to 7 mdB. If both numbers were measured in a like-for-like way, that is a large reduction, and “nearly threefold” is a fair description of the arithmetic. What cannot be checked from this week’s materials is whether a third party would record the same 7 mdB, whether the 2025 and 2026 tests used the same method, or whether the unnamed “closest competitor’s most recent published result” is the right baseline. Those are exactly the points on which a paper, or a measurement by someone other than the two vendors, would help.

The production claim should be read in the same light. “Ready to ramp to production” is a vendor’s judgement about manufacturing readiness, supported in the release by a reference to a high-volume line, yield and repeatability. It is not a shipment figure, a customer qualification, or an independent process-audit. Xanadu’s earlier provisional DARPA quantum-benchmarking award already placed the company in a public programme for assessing quantum hardware; this week’s switch result is a separate, company-run benchmark, not a DARPA score.

The wider photonic market is also starting to talk about optical switches as more than a quantum-only part. PsiQuantum has described turning its quantum photonics into an optical circuit switch for AI, a different company making a different claim about the same class of component. Other groups, including a Jülich-led team reporting very fast qubit gates, are likewise publishing hardware performance figures that have to be read as the authors’ own measurements until they are reproduced. The Xanadu optical switch announcement belongs in that category: a sharp, specific, vendor-reported number, with a clear prior baseline, and without a third-party test attached.

The next facts that would change the reading are a paper or documented method for the 7 mdB result, a named competitor comparison that can be checked, and evidence of switches actually moving into production rather than being described as ready to do so. Until then, the 7 mdB figure should be reported as Xanadu and HyperLight’s benchmark, not as a settled industry measurement.