2 March 2026

 

On February 26th, 2026, Canadian quantum computing company Xanadu and global defence and technology group Lockheed Martin announced a joint research initiative focused on the foundations of quantum machine learning (QML). Rather than promising near-term breakthroughs or commercial applications, the collaboration deliberately targets unresolved theoretical questions about how learning might work on future quantum computers.

The partnership centres on generative models and so-called quantum-native operations — mathematical processes that arise naturally in quantum physics but are difficult to reproduce efficiently on classical computers. The aim is not to build products, but to better understand whether quantum systems could one day process and model information in ways that classical machine learning cannot.

Why quantum machine learning is still an open question

Machine learning underpins much of today’s digital infrastructure, from recommendation engines to generative AI systems. Many of these systems rely on generative models that learn statistical patterns from large datasets and then produce new data that resembles what they were trained on.

These models are effective, but they are also resource-intensive. Training them often requires enormous datasets, specialised hardware, and significant energy consumption. This has prompted researchers to ask whether alternative computing paradigms could offer different trade-offs.

Quantum machine learning explores that question using quantum computers, which operate on fundamentally different principles from classical machines. Instead of binary bits that are either 0 or 1, quantum computers use qubits, which can exist in combinations of states at the same time and become correlated through quantum effects such as entanglement.

In theory, this allows quantum systems to represent and manipulate information in richer ways. In practice, quantum computers remain small, fragile, and difficult to scale. Whether they can deliver meaningful advantages for machine learning remains an open research problem — one that this collaboration explicitly acknowledges.

What this research is actually about

The Xanadu–Lockheed Martin initiative focuses on foundational theory, not applications. The research will examine how basic elements of quantum computation — often referred to as quantum primitives — could be used to construct new types of generative models.

A key area of interest is the role of Fourier-based operations, mathematical transformations that are central to signal processing and physics. In quantum systems, these operations arise naturally and can be executed very differently than on classical hardware. The researchers want to understand whether such operations could enable models that represent complex data structures more efficiently, potentially using fewer parameters or less training data.

Christian Weedbrook, founder and CEO of Xanadu, framed the work as a reassessment of assumptions rather than a path to quick wins:

“If quantum computers are ever going to offer meaningful advantages for learning, we need to understand how learning itself should be defined in a quantum setting.”

A deliberate absence of short-term promises

One of the more notable aspects of the announcement is what it does not claim. There are no timelines for deployment, no benchmarks against classical AI, and no assurances of quantum advantage. This restraint is significant in a field that has often struggled with inflated expectations.

Lockheed Martin’s involvement reflects a long-horizon view of emerging technologies. The company has previously invested in quantum sensing, communications, and computing as part of its advanced research portfolio. According to Dani Couger, Quantum Technologies Lead at Lockheed Martin, the collaboration is about building understanding rather than capability in the near term:

“This work helps us explore how future quantum systems might support advanced computation and sensing — not what they can deliver today.”

Xanadu’s position in the quantum landscape

Founded in 2016, Xanadu has positioned itself as a full-stack quantum computing company, working across hardware, software, and theory. Its hardware efforts focus on photonic quantum computing, which uses particles of light rather than superconducting circuits. Photonics is often cited as a promising approach because it aligns more naturally with existing optical technologies, though it faces its own engineering challenges.

On the software side, Xanadu develops PennyLane, an open-source library designed to connect quantum computing with machine learning workflows. PennyLane is widely used in research settings, particularly for experimenting with hybrid models that combine classical and quantum components.

This collaboration fits with Xanadu’s broader emphasis on open research and theoretical rigour, reinforcing its role as a contributor to the scientific foundations of quantum computing rather than a company focused solely on hardware milestones.

Potential relevance — eventually

The announcement points to possible long-term application areas such as defence, finance, and pharmaceuticals — sectors where modelling complex systems with limited or sensitive data is a recurring challenge. However, these references are speculative and framed as future possibilities, not near-term use cases.

That distinction matters. Quantum machine learning remains an experimental field, and there is no consensus that quantum systems will outperform classical machine learning for real-world problems. What collaborations like this offer instead is a way to clarify where quantum approaches might make sense — and where they likely will not.

A step back that may be necessary

In an industry often driven by roadmaps and performance claims, the Xanadu–Lockheed Martin partnership stands out for its focus on first principles. By stepping back to examine how learning, representation, and computation might work in a quantum context, the initiative acknowledges a basic reality: useful quantum machine learning will not emerge without a deeper theoretical foundation.

Whether this work will eventually translate into practical systems remains to be seen. But as quantum computing continues to move — slowly — from promise to practice, efforts like this help separate what is plausible from what is merely aspirational.

 

Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks. 

👉 https://buymeacoffee.com/movetheneedle.news