Xanadu

Overview

Xanadu is a Toronto-based full-stack photonic quantum computing company founded in 2016, pursuing a hardware-software strategy built on continuous-variable (CV) and Gaussian boson sampling (GBS) photonic architectures. Unlike superconducting or trapped-ion approaches, Xanadu's photons operate at room temperature in principle, though its current systems require cryogenic components for certain detectors. The company's core thesis is that photonics offers a scalable path to fault-tolerant quantum computing via a modular, networked architecture — connecting smaller photonic chips via optical fiber to build larger logical systems, a strategy the company calls 'photonic interconnects.' This positions Xanadu as a longer-term architectural bet relative to superconducting players, with the tradeoff being that near-term qubit counts and gate fidelities remain lower than leading superconducting systems.

Xanadu's commercial strategy operates on two tracks. First, it develops and maintains PennyLane, an open-source quantum machine learning (QML) software framework that has achieved significant adoption across the quantum computing community, functioning as both a developer acquisition tool and a hardware-agnostic platform that runs on IBM, Google, Amazon Braket, and Xanadu's own hardware. PennyLane is widely regarded as one of the most actively developed QML libraries, with a large contributor community. Second, Xanadu operates Borealis, its cloud-accessible photonic quantum processing unit (QPU), which became notable in June 2022 when the company published a Nature paper demonstrating quantum computational advantage — performing a Gaussian boson sampling task in 36 microseconds that would take the best classical algorithms an estimated 9,000 years on a supercomputer.

Xanadu's commercial partnerships have historically included cloud access agreements via Amazon Braket, where Borealis is accessible to AWS customers, broadening its reach without requiring direct enterprise sales infrastructure. The company has also engaged with research institutions and government bodies, particularly in Canada under the National Quantum Strategy framework. Revenue streams are likely a combination of cloud QPU access fees, software licensing or services around PennyLane, and research grants, though the company has not historically disclosed detailed revenue figures. As of early 2026, Xanadu reportedly completed a public listing on Nasdaq at a $3.1 billion valuation, which if confirmed would make it one of the highest-valued photonic quantum companies to reach public markets.

In the competitive landscape, Xanadu occupies a distinctive but challenging position. It is the most prominent pure-play photonic quantum computing company alongside PsiQuantum (which remains private and focused on silicon photonics at much larger scale). It competes indirectly with superconducting leaders IBM, Google, and IonQ's trapped-ion systems for cloud QPU mindshare and developer ecosystem. Its PennyLane software is a genuine differentiator and a strategic moat — but hardware differentiation is harder to sustain against well-capitalized incumbents. Xanadu's bet is that photonic interconnects will enable scalable fault-tolerant QC before superconducting systems can solve their connectivity and crosstalk challenges at scale.

Leadership

Technology

Xanadu's hardware is built on photonic integrated circuits (PICs) that encode quantum information in the quadratures of the electromagnetic field — a continuous-variable (CV) approach — rather than discrete qubit states. Its near-term systems use Gaussian boson sampling, where squeezed light states are injected into a reconfigurable linear optical interferometer and measured with photon-number-resolving detectors. The Borealis system, announced in 2022, features 216 squeezed modes and a fully programmable architecture using loop-based time-multiplexing, making it the first photonic quantum advantage demonstration on a programmable device. This is technically significant because prior photonic advantage claims (e.g., USTC's Jiuzhang) used fixed, non-programmable optical circuits.