When Light Meets Quantum

In part one I argued quantum is having its transistor moment. In part two I showed you photonics, the trillion dollar light industry holding up the AI boom. Here is where the two become one.

Start with a number that surprised me. Xanadu, the Toronto based photonic quantum company, went public on March 27, 2026, the first pure play photonic quantum computing company to list on a major exchange, trading on Nasdaq and the TSX under XNDU. The transaction materials described a pro forma enterprise value of about $3.1 billion and a market capitalization near $3.6 billion (it's currently $4.1 billion). That is a public market vote on the idea that the qubit of the future might be made of light.

This is the post where the two industries I have been describing turn out to be the same industry. The photonics railroad from part two, all those foundries and packaging lines and 300 millimeter wafers, is exactly the track the photonic quantum industry is now trying to ride. If light is the new copper for AI, it may also be the new copper for quantum.

Why photons make decent qubits

Every qubit modality has a trade off. The photonic bet is that it is worth it. Photons, unlike superconducting qubits, do not need to sit inside a dilution refrigerator colder than deep space. They can, in principle, operate at room temperature. They do not couple strongly to their environment, which is another way of saying they are slow to decohere. And, critically for the business case, they travel through the same optical fiber that already carries the world's telecom traffic, and they can be fabricated on the same semiconductor foundry lines that already make silicon photonics.

That last point is the whole argument. If you can build a million qubit quantum computer on a standard 300 millimeter wafer line, using the same tools the photonics industry already paid for, you have changed the economics of quantum scaling. You are no longer building one exotic machine at a time. You are running a wafer fab.

Two architectures, one light beam

There are two main ways to turn a photon into a qubit, and the leaders have taken different sides.

The discrete variable approach encodes information in single photons, in properties like polarization or path. The computation runs through linear optical elements, beam splitters and phase shifters, plus measurements, in a scheme called measurement based or cluster state computing. The aggressive exponent is PsiQuantum, which uses fusion based computing to stitch together large entangled states from smaller ones. In February 2025 PsiQuantum announced Omega, a quantum photonic chipset it describes as mass manufacturable and purpose built for million qubit scale systems, published in Nature. The chips are made at GlobalFoundries' Fab 8 in New York. In September 2025 it closed a Series E of $1 billion at a $7 billion valuation, led by BlackRock, Temasek and Baillie Gifford, and is building utility scale facilities in Chicago and Moreton Bay, Australia, the latter backed by A$940 million in Australian government funding.

The continuous variable approach is Xanadu's bet. It encodes information in the continuous properties of squeezed light, using a code called GKP encoding, and computes by measuring the light. In early 2025 Xanadu announced Aurora, the world's first modular, networked photonic quantum computer with real time error correction, detailed in Nature. Aurora demonstrated 12 logical GKP qubits with real time error correction, and Xanadu reported cutting optical loss by 60 percent over the year, part of a 20 fold improvement over three years. In June 2025 the team showed the first on chip generation of optical GKP states on a 300 millimeter wafer platform. Its open source software stack, PennyLane, is used by about 47 percent of quantum developers. A public listing on top of that is a company telling the market it has a manufacturable path.

Around these two are the rest of the field. Quandela, the French company, launched Belenos in 2025, a 12 qubit photonic machine it says delivers 4,000 times the power of its previous generation, and has been integrated into the EuroHPC supercomputing infrastructure. In June 2026, Quandela validated a low latency integration path between its photonic processors and Nvidia's AI infrastructure, a real signal that photonic quantum is being treated as an accelerator that sits next to classical AI. QuiX, based in Twente in the Netherlands, calls itself the market leader in quantum photonic processors and has sold hardware to Quandela. Photonic Inc., the Vancouver based company spun out of Simon Fraser University, builds its qubits from silicon color centers optically linked by photons, in a long running partnership with Microsoft aimed at distributed quantum computing and networking.

And photonics is not only a way to make qubits. It is also how you control and read out the non photonic ones, the superconducting and trapped ion machines. The control lasers, the optical interconnects and the readout are all photonics work. Photonics shows up in every quantum lab, not just the photonic ones.

The shared infrastructure, and the shared problem

The reason a photonic quantum computer is plausible at all is that it does not have to invent its supply chain. It inherits the one from part two. The same PIC foundries, the same wafer level packaging, the same fiber and detector ecosystem. Xanadu fabricates Aurora's circuits at its own facility and partnered with Thorlabs for fiber and EV Group for bonding. PsiQuantum runs through GlobalFoundries. Quandela uses partner foundries for its photonic integrated circuits. These are the same rails.

The shared problem is photon loss. Photons are great qubits until they disappear, and in a long computation you need a lot of them, generated, routed and detected, with very low loss. Xanadu's 60 percent reduction in optical loss is not a footnote, it is the thing that determines whether the architecture scales. High quality single photon sources, efficient detectors and the probabilistic nature of some photonic operations are the open problems. They are not physics hard anymore, they are manufacturing hard. That is the transistor moment again.

The Israeli thread

A photonic quantum computer is a machine controlled by lasers, modulators and fast electronics, run by a software stack that compiles quantum programs into optical operations. Quantum Machines builds the control electronics. Classiq builds the compilation software. The Israeli Quantum Computing Center, run with the Israel Innovation Authority, is where a lot of this gets integrated and stress tested. Israel's bet, that the value is in the control and software layer rather than only in the qubit, holds up regardless of whether the winning qubit is a superconducting loop, a trapped ion or a photon. If light wins, the Israeli companies that already control and compile for quantum machines are positioned to do it for the photonic ones too.

Beyond computing, the quantum internet

Photonic quantum is also the natural substrate for what comes after computing. Quantum key distribution uses single photons to share encryption keys in a way that is provably secure against eavesdropping, and it runs on fiber. Quantum networks, the long term vision of a quantum internet that links processors over distance, require photons because they are the only qubit that travels. Quantum sensing and quantum random number generation lean on the same stack. The light revolution is not just about a faster computer. It is about building the quantum version of the network.

The next wave of compute

Quantum is real and accelerating, but still early. Photonics is mature, enormous and growing faster because of AI. And the most credible path to a scalable quantum machine runs straight through the photonics supply chain that already exists. PsiQuantum and Xanadu are not building quantum computers from scratch. They are building them on a railroad the photonics industry already laid, paid for and tooled up. If they succeed, the trillion dollar light industry does not just gain a new customer. It becomes the substrate of the next computing era.

My takeaway for anyone watching this space, whether you are an investor, an engineer or a founder, is to stop treating photonics and quantum as two sectors. They are one stack, and the companies that own the layers where they overlap, the foundries, the packaging, the control electronics, the software and the fiber, are positioned to win no matter which qubit carries the day. Light is the new copper. Quantum is about to ride it.