By XanaduLearning Objectives
- Understand what makes Xanadu's photonic approach to quantum computing distinct from other quantum hardware.
- Learn how the Aurora and Borealis systems are built and how their specifications compare.
- See how PennyLane, Xanadu's open-source framework, lets you build hybrid quantum machine learning models today.
What Is Xanadu?
Xanadu is a quantum computing company founded in 2016 by Christian Weedbrook and headquartered in Toronto, Canada. The company went public on March 27, 2026, dual-listed on Nasdaq and the Toronto Stock Exchange under the ticker XNDU.
What sets Xanadu apart is its choice of technology. Instead of cooling tiny superconducting circuits to temperatures colder than deep space, Xanadu uses photonic quantum computing — encoding quantum information in particles of light. This approach has two practical advantages that matter for anyone trying to scale a quantum computer: it operates at or near room temperature, and the systems are naturally networkable over standard optical fiber.
For learners on this platform, the most important thing about Xanadu is not the hardware at all — it is PennyLane, the company's open-source software framework and the leading tool for quantum machine learning. PennyLane is free, runs on your own laptop, and connects to the same machine learning libraries you may already know. We will build up to it.
Photonic Quantum Computing
💡Key Concept
Most quantum computers store information in superconducting circuits or trapped ions that must be chilled to near absolute zero. Xanadu instead encodes quantum information in squeezed states of light — carefully prepared beams of photons whose quantum noise has been "squeezed" below the normal limit. Because light travels well at ordinary temperatures and moves naturally through optical fiber, photonic machines can run at or near room temperature and can be linked together over fiber the same way data centers are networked. That combination — no extreme cooling, and a clear path to connecting many modules — is the central bet behind Xanadu's design.
The room-temperature property removes one of the biggest engineering burdens in the field: the bulky, power-hungry refrigeration that other architectures depend on. The networkability matters even more for the long term, because scaling a quantum computer is largely a problem of connecting many smaller pieces into one coherent system. Optical fiber gives Xanadu a ready-made way to do that.
Aurora and Borealis
Xanadu has two flagship systems, each illustrating a different part of its strategy.
Aurora, announced in January 2025 and published in the journal Nature, is described as the first modular, networked, and scalable photonic quantum computer. It is built from 35 photonic chips linked together by 13 kilometers of optical fiber — a concrete demonstration of the "connect many modules over fiber" approach.
Borealis is a 216-qubit photonic system. In June 2026 it received public cloud access through both Xanadu Cloud and Amazon Web Services Braket, meaning researchers can now run programs on real photonic hardware without owning any of it.
| System | Type | Scale | Notable detail |
|---|---|---|---|
| Aurora | Modular photonic quantum computer | 35 photonic chips | Linked by 13 kilometers of optical fiber; published in Nature (January 2025) |
| Borealis | Photonic quantum system | 216 qubits | Public cloud access via Xanadu Cloud and AWS Braket (June 2026) |
The AI Angle
This is where Xanadu becomes a tool you can actually use. PennyLane is Xanadu's open-source software framework and the leading tool for quantum machine learning — the field that combines quantum circuits with the training methods used in ordinary machine learning.
PennyLane is cross-compatible with PyTorch, TensorFlow, and JAX. That means you can drop a quantum circuit into a model built with the same libraries that power modern AI, and PennyLane handles the connection between them. Crucially, it supports automatic differentiation across quantum circuits, so you can train a hybrid quantum-classical model end to end using the same gradient-based optimization that trains a neural network.
In plain terms: PennyLane lets a researcher treat a quantum circuit as one more trainable layer inside a familiar AI workflow. You do not need access to quantum hardware to start — PennyLane runs on simulators on your own machine, and when you are ready, the same code can target real photonic devices through the cloud.
How You Access It
| Tool | Best For |
|---|---|
| PennyLane | Free, open-source quantum machine learning framework — runs on your laptop, cross-compatible with PyTorch, TensorFlow, and JAX |
| Xanadu Cloud | Cloud access to Xanadu's photonic hardware, including the Borealis system |
| Borealis on AWS Braket | Run the 216-qubit Borealis photonic system through Amazon Web Services Braket |
Strengths
- PennyLane is genuinely usable today — open-source, free, and installable with a single command, with no quantum hardware required to begin.
- Fits existing AI workflows through cross-compatibility with PyTorch, TensorFlow, and JAX, plus automatic differentiation across quantum circuits.
- Room-temperature photonic hardware avoids the costly cryogenic cooling that other quantum architectures require.
- Naturally networkable over optical fiber, giving a clear architectural path toward scaling many modules into one system.
- Real cloud access to photonic hardware via Xanadu Cloud and AWS Braket, so research is not limited to simulation.
Limitations & Considerations
- Quantum machine learning is still an early research field; clear, practical advantages over classical methods are not yet established for most tasks.
- Photonic quantum computing competes with more mature superconducting and trapped-ion approaches, and the best architecture for the long run remains an open question.
- Running on real hardware introduces noise and error rates that simulators do not show, so results can differ from idealized expectations.
- Cloud time on photonic systems can be limited and may carry cost, so most early experimentation happens on local simulators.
Best Use Cases
| Scenario | Why Xanadu fits |
|---|---|
| Learning quantum machine learning hands-on | PennyLane is free, runs locally, and uses familiar PyTorch, TensorFlow, or JAX workflows |
| Building hybrid quantum-classical models | Automatic differentiation lets you train quantum circuits like any other model layer |
| Research needing real photonic hardware | Borealis is reachable through Xanadu Cloud and AWS Braket |
| Exploring room-temperature, networkable quantum designs | Xanadu's photonic architecture is purpose-built around fiber networking |
Getting Started
- Install PennyLane from its open-source distribution and run it on your own machine — no quantum hardware needed.
- Work through PennyLane's introductory tutorials to build your first quantum circuit as a trainable component.
- Connect a circuit to a model in PyTorch, TensorFlow, or JAX and train it end to end using automatic differentiation.
- When you are ready for real hardware, target the Borealis photonic system through Xanadu Cloud or AWS Braket.
Key Takeaways
- Xanadu is a Toronto-based, publicly traded company building photonic quantum computers that run at near room temperature and connect over optical fiber.
- Aurora demonstrated a modular, networked design using 35 photonic chips and 13 kilometers of fiber; Borealis is a 216-qubit system now available in the cloud.
- PennyLane is the real AI hook — the leading open-source quantum machine learning framework, cross-compatible with PyTorch, TensorFlow, and JAX.
- Because PennyLane supports automatic differentiation across quantum circuits, you can train hybrid quantum-classical models with the same methods used for neural networks.
- You can start for free today on simulators and graduate to real photonic hardware through Xanadu Cloud or AWS Braket when ready.
