Photonic Inc.

(This profile is one entry in my 2025 series on quantum hardware roadmaps and CRQC risk. For the cross‑vendor overview, filters, and links to all companies, see Quantum Hardware Companies and Roadmaps Comparison 2025.)

(Updated 25 September 2026. This revision adds Photonic’s 2025 and 2026 funding, leadership changes, government programmes, the TELUS fibre demonstration, the peer-reviewed SHYPS paper, the company’s distributed resource-estimation work and Project VANGUARD)

Introduction

Photonic Inc. is a quantum computing company based in Coquitlam, near Vancouver, building a distributed, fault-tolerant quantum computer from silicon spin qubits that are linked by photons. The qubits are “T centres”, defects in silicon made of two carbon atoms and one hydrogen atom, which emit and absorb single photons at about 1326 nm in the telecom O-band. Every qubit therefore has an optical interface, and qubits in separate chips, cryostats or buildings can be entangled over standard telecom fibre. Photonic calls the design “Entanglement First™” and built networking into the computer from the start, combining quantum computing and quantum communication in one platform.

Dr. Stephanie Simmons, a silicon quantum physicist at Simon Fraser University (SFU), founded Photonic in 2016 and is its Chief Quantum Officer. The leadership team changed in 2026 as the company prepared to commercialize. Don Mattrick became CEO in March 2026. He co-founded Distinctive Software (later EA Vancouver), ran Microsoft’s interactive entertainment business and was CEO of Zynga, and he was one of Photonic’s first investors. Dr. Paul Terry, CEO since 2019, moved to Chief Product Officer and is titled Chief Engineering Officer in the company’s September 2026 releases. Alex van Someren, founder of nCipher and a former UK Chief Scientific Adviser for National Security, became Executive Chair in February 2026. In July the company hired Orlagh Neary as Chief Marketing Officer and Briony Shipman as Vice President of Global Government Affairs.

Photonic had raised about US$140 million by late 2023, with Microsoft and the British Columbia Investment Management Corporation (BCI) among its backers. Microsoft is both an investor and a partner, with plans to offer Photonic hardware through Azure. A new round opened with a CAD$180 million first close in January 2026, led by Planet First Partners, with RBC and TELUS joining and BCI and Microsoft returning. It closed in May 2026 at more than US$200 million (CAD$275 million) and a US$2 billion post-money valuation. The final close added BDC’s defence-focused StrongNorth Fund, Export Development Canada, Bell Ventures, InBC, Firgun Ventures and Mubadala Capital, and brought total funding to more than US$350 million. The company reports a team of more than 180 people, with operations in the United States and the United Kingdom.

Governments on both sides of the border have now reviewed the approach. In November 2025, Photonic was one of 11 companies selected by DARPA’s Quantum Benchmarking Initiative (QBI) to advance to Stage B after proposing a “utility-scale” quantum computer concept in Stage A. QBI’s goal is to verify whether any approach can reach an industrially useful, fault-tolerant quantum computer, one whose computational value exceeds its cost, by 2033. A month later, Canada selected Photonic for the first phase of its Quantum Champions Program, with up to CAD$23 million.

In my assessment, Photonic is now one of the best-funded and most closely government-reviewed quantum hardware companies that has not yet shipped a commercial system, and it has still not shown a logical qubit in hardware. In 2023, Dr. Simmons said Photonic expected to offer a distributed, fault-tolerant quantum computing solution within five years, which points to about 2028. What the company has demonstrated so far is entanglement and a teleported gate between two modules (2024), teleportation of quantum information into a matter qubit over 30 km of deployed TELUS fibre (2026, announced but not yet written up), and peer-reviewed simulations of its error-correcting codes (2026). It has not published a physical-qubit count for any system, a logical-qubit demonstration or a system-level benchmark.

Milestones & Roadmap

Photonic’s main milestones, oldest first:

• 2020 – T-Centre Spin-Photon Properties Characterized: Simmons’ group at SFU reported in PRX Quantum that T centres in isotopically purified silicon-28 combine long-lived electron and nuclear spin states with optical transitions near 1326 nm in the telecom O-band, a wavelength well suited to low-loss fibre. The defect consists of two carbon atoms and one hydrogen atom. These measurements were made on ensembles of T centres; single centres followed two years later.
• 2022 – Single T Centres in Silicon Photonic Devices: The team reported in Nature the first optical observation of single spins in silicon. They integrated individually addressable T centres into silicon photonic structures and measured their spin-dependent telecom-band optical transitions. Several authors listed dual affiliations with SFU and Photonic. With this paper, the group established that CMOS-compatible processes can host individually addressable spin-photon qubits, the basic building block of Photonic’s modules.
• 2023 – Architecture Unveiled & Funding: Photonic emerged from stealth in late 2023, announcing its “Entanglement First™” architecture for a networked quantum computer and a $100 M round that brought total funding to $140 M. The company described its plan as “one of the world’s first scalable, fault-tolerant, and unified quantum computing and networking platforms” based on photonically linked silicon spin qubits. Dr. Simmons said Photonic expected to offer a distributed, fault-tolerant quantum solution within five years. In November 2023, Photonic and Microsoft announced a strategic collaboration to co-innovate on quantum networking and to integrate Photonic’s hardware into Azure as it matures.
• May 2024 – Distributed Entanglement Demonstrated: Photonic entangled T-centre qubits in two separate modules, each in its own cryostat and connected by telecom fibre, and consumed that entanglement to run a teleported CNOT gate sequence between them. Most quantum processors entangle qubits only on the same chip; this was one of the first demonstrations of entanglement between silicon-based modules. Microsoft’s quantum team described it as a fundamental capability on the path to networked quantum computing. The accompanying arXiv paper calls the results “preliminary demonstrations” and a proof of concept for T centres as a distributed computing and networking platform. As of September 2026, the arXiv record still shows only the original June 2024 version and no journal reference.
• Feb 2025 – SHYPS QLDPC Codes: Photonic released a preprint on a new family of Quantum Low-Density Parity Check (QLDPC) codes it calls SHYPS (Subsystem Hypergraph Product Simplex). Earlier QLDPC work had focused mostly on storing quantum information; SHYPS was designed to run logical operations efficiently inside the code. Photonic’s announcement cited up to 20× fewer physical qubits than surface codes, and an earlier version of this profile repeated that figure. The peer-reviewed version (May 2026, below) reports 2× and 3.5× fewer physical qubits than comparable surface-code encodings at the code sizes the authors simulated.
• Jun 2025 – UK R&D Facility: Photonic announced a UK research and development facility, with more than £25 million of investment over three years and more than 30 jobs.
• Sep 2025 – Electrically Triggered Spin-Photon Device in Silicon (Nature Photonics): SFU and Photonic Inc. reported the first electrically injected single-photon source in silicon using T-centre defects, integrated in nanophotonic p-i-n diode structures. The work also demonstrated heralded spin initialization (~92% fidelity) via detected telecom-band photons, showing that T-centre spin qubits can be driven and prepared using standard electronic pulses (not only lasers). This advances CMOS-compatible, parallelizable qubit control and on-chip light sources for scalable, fiber-ready quantum processors.
• Nov 2025 – DARPA QBI Stage B: In November 2025, Photonic was advanced to Stage B of DARPA’s Quantum Benchmarking Initiative, after its Stage A concept for a Utility-Scale Quantum Computer passed review. Stage B is a year-long examination of each company’s research and development plan, risks and prototypes, and Canadian press reported funding of up to US$15 million per company for this stage. Companies that pass move to Stage C, where independent government teams test whether the design can be built and operated as intended.
• Dec 2025 – Canadian Quantum Champions Program: Canada selected Photonic for Phase 1 of its Quantum Champions Program, with up to CAD$23 million. Innovation, Science and Economic Development Canada runs the programme, with technical benchmarking by the National Research Council. Phase 1 funds four companies: Anyon, Nord Quantique, Photonic and Xanadu.
• Dec 2025 – Distributed Resource Estimation: At the Q2B conference, Photonic published “SHYPS to Shor’s: A Call for Distributed QRE”, a whitepaper with a resource estimate for Shor’s algorithm on a fully distributed SHYPS architecture that counts the cost of operations within and between modules. Photonic argues that published Shor’s estimates assume a single monolithic machine and so understate what a real machine will cost. The qubit and runtime figures are in the whitepaper download, not in the public release, and I have not yet reproduced them.
• Jan 2026 – New Funding Round Opens: A CAD$180 million (US$130 million) first close led by Planet First Partners, with RBC and TELUS as new investors and BCI and Microsoft returning.
• Feb 2026 – Teleportation over TELUS Fibre: Photonic and TELUS announced the teleportation of quantum information over 30 km of TELUS’s installed commercial fibre into a matter-based quantum processor that can store and use the information. The companies say earlier field tests of this kind ended with the information in photons, not in a stored qubit. They also expanded their partnership to cover work from quantum data centres to nationwide encrypted networks. As of September 2026, I have found no paper or published fidelity figures for this result. The same month, Alex van Someren became Executive Chair.
• Mar 2026 – New CEO: Don Mattrick succeeded Dr. Paul Terry as CEO, with Terry moving to Chief Product Officer. Photonic described the change as sharpening its commercialization focus.
• May 2026 – Round Closes at a US$2 Billion Valuation: The final close brought the round to more than US$200 million, with BDC’s StrongNorth Fund, Export Development Canada, Bell Ventures, InBC, Firgun Ventures and Mubadala Capital joining.
• May–Aug 2026 – SHYPS Peer-Reviewed: “Computing efficiently in QLDPC codes” was published in Nature Communications (online 20 May 2026; Photonic announced it on 25 August). In circuit-level simulations, the [49, 9, 4] SHYPS code outperformed nine distance-3 rotated surface codes while using half the physical qubits, and the [225, 16, 8] code performed comparably to sixteen distance-7 surface codes with 3.5× fewer qubits. The authors released their simulation code and data on GitHub and Zenodo.
• Jul 2026 – Commercial Hires: Orlagh Neary joined as Chief Marketing Officer and Briony Shipman as Vice President of Global Government Affairs, which Photonic presented as adding commercial and policy depth.
• Sep 2026 – Project VANGUARD: Photonic proposed a multi-tenant semiconductor fabrication and packaging facility in Canada costing up to CAD$500 million, serving quantum, AI, aerospace, defence and sensing customers. The proposal was listed in the prospectus for the Canada Investment Summit in Toronto. It is a proposal, not a funded project.
• Sep 2026 – Resource Estimation with Microsoft: Photonic and Microsoft announced a collaboration to model distributed architectures and SHYPS-type codes in Microsoft’s open-source Quantum Resource Estimator, so that developers can estimate qubit counts, runtime and overhead for algorithms on networked machines. The capabilities have not yet been released and no release date has been given.
• Next: DARPA’s Stage B review runs for about a year from November 2025, and a Stage C decision would be the next external test of Photonic’s plan. The technical milestones to watch are more direct: published inter-module entanglement rates and fidelities, modules with more T-centre qubits than the few used in experiments so far, and a first logical qubit encoded in a QLDPC code on T-centre hardware. Photonic says customers will eventually reach its machines through Microsoft Azure or dedicated private systems; it has not given a date.

Focus on Fault Tolerance

Photonic has designed its architecture around fault tolerance from the start. Its founders argue that a useful quantum computer needs error correction built in early, and that networking many modules is the only realistic way past the physical limits of a single device. “Entanglement First” names that priority: high-quality entanglement between qubits, both locally and across modules, gives the machine the connectivity that efficient error-correcting codes need. The SHYPS codes are the clearest example. Like other QLDPC codes, they use non-local parity checks, connections between qubits that are physically far apart, which a surface code on a nearest-neighbour grid avoids. Photonic’s optical links can in principle supply those connections within and between modules, which is why qubit connectivity is one of the capabilities I track separately.

The peer-reviewed SHYPS paper gives the most concrete picture so far of what the codes offer. In circuit-level simulations, the authors found SHYPS codes competitive with surface codes while using 2× to 3.5× fewer physical qubits at the sizes tested. The codes can also compile any Clifford operation, the class of quantum logic that excludes the costly T gates, in a depth that does not grow with code distance. For a distance-20 code, the authors estimate that a logical CNOT would take 4 time steps with SHYPS against roughly 40 with lattice surgery on surface codes, a tenfold gain in logical clock speed. They also simulated a depth-126 logical circuit on 18 logical qubits and measured error rates close to those of an idle quantum memory. The authors disclose four patent applications covering parts of the work.

The same paper records the limits. The larger SHYPS code has a lower pseudo-threshold than the surface code it is compared with, about 0.35% against about 0.8%, and it only outperforms that surface code once physical error rates fall below 0.05%. All of the results are simulations under a uniform circuit-level noise model, and Photonic has not published T-centre hardware error rates in that range. The T gates that dominate algorithms such as Shor’s still rely on known magic-state methods that the paper does not re-cost. The authors write that, for any architecture able to provide the connectivity, QLDPC codes now look like the stronger path, and that connectivity and simplicity are the remaining reasons to choose the surface code.

Photonic has also demonstrated some of the hardware ingredients for fault tolerance. In its 2025 Nature Photonics work with SFU, the team used a detected telecom photon to herald the spin initialization of a T centre with about 92% fidelity. Heralding tells the controller when an operation has succeeded, so only successful instances are used, a standard technique in networked error correction. Photonic has also reported work on tuning the optical frequencies of T centres, which is needed to make photons from different centres interfere reliably.

The next hardware steps would be to protect a single logical qubit, then to operate small networks of logical qubits across modules. Neither has been shown publicly. DARPA’s Stage B review is examining exactly this path, and the September 2026 resource-estimation collaboration with Microsoft is intended to let outsiders model what distributed SHYPS machines would need for real algorithms. Until those models are released, nobody outside Photonic can reproduce its system-level numbers.

Photonic’s bet is that distributed quantum computing, many modules of qubits entangled together, is the only realistic route to the scale and reliability that useful algorithms need. As then-CEO Paul Terry put it: “All large-scale systems are fundamentally networks… At Photonic, we are building networks of qubits to create infinitely scalable quantum computers.” If the hardware reaches the error rates the codes require, that bet would give Photonic an efficiency advantage over nearest-neighbour architectures. Whether it will is an experimental question the company has not yet answered in public.

CRQC Implications

Q-Day, the day a quantum computer can break today’s public-key cryptography such as RSA-2048, is the benchmark for a cryptographically relevant quantum computer (CRQC). Photonic markets its work for materials science, drug discovery, climate and security, but a machine that reaches its design goal would also be able to run Shor’s algorithm. Photonic acknowledges this itself: its first public resource estimate, in December 2025, was for Shor’s algorithm.

Published estimates for breaking RSA-2048 have fallen sharply. In May 2025, Craig Gidney of Google Quantum AI estimated that fewer than one million noisy qubits could factor a 2048-bit RSA key in under a week, using surface codes on a single large machine. A March 2026 preprint on neutral-atom hardware used high-rate codes on reconfigurable atom arrays to estimate about 102,000 physical qubits and roughly 97 days for RSA-2048, and its title claims Shor’s algorithm is possible with as few as 10,000 atomic qubits at lower parallelism. The neutral-atom result relies on the same lever Photonic pulls: more connectivity allows more efficient codes, and more efficient codes cut the physical-qubit count.

Photonic’s December 2025 whitepaper argues that monolithic estimates like Gidney’s understate the real cost, because no qubit modality will easily fit a million qubits into one machine, and the links between modules add overhead of their own. Other researchers have started to model this: an April 2025 preprint on network requirements for distributed quantum computation compared monolithic and distributed resource estimates for factoring RSA-2048. Photonic has not put its own Shor’s figures in its public releases, and until its models are available in Microsoft’s estimator I cannot check them. The runtime of a distributed CRQC will depend on how fast modules can share entanglement. Photonic’s stated target is about 200 kHz at 99.8% fidelity, and it has not published inter-module rates near that figure.

Photonic’s demonstrated hardware is still at the scale of a few qubits. As I put it previously, its electrically triggered spin-photon result was a hardware milestone that clears another hurdle toward a scalable quantum computer, the kind of innovation needed to eventually reach the huge number of qubits required for breaking encryption. But entangling two modules and publishing more efficient codes are pieces of the puzzle, not a cryptanalysis machine. Operating a very large cryogenic system reliably, raising qubit fidelities, and running error-correction cycles in hardware all remain to be done at scale.

From a security viewpoint, Photonic is a company to keep on the radar. Its 2023 target of a fault-tolerant system within five years would, if met, put an early CRQC-class capability on the table before the end of the decade. Defence-linked funders back it on both sides of the border, including DARPA through QBI and BDC’s defence-focused StrongNorth Fund in Canada. A large Photonic machine could be used to attack cryptography just as well as to simulate molecules, and because its qubits communicate at telecom wavelengths, such a machine could in principle be spread across several sites connected by existing fibre.

For now, Photonic’s advances have not changed the Q-Day timeline. Three results would change my view of how fast its path could move: a logical qubit on T-centre hardware, published inter-module entanglement rates approaching the company’s own 200 kHz target, and a distributed Shor’s estimate that someone outside Photonic can reproduce.

Modality & Strengths/Trade-offs

Photonic’s qubit modality is a hybrid of solid-state spins and photonics, a branch of the broader silicon spin qubit family. Each qubit is a T-centre defect in silicon-28, an isotope without nuclear spin that gives the defect a quiet magnetic environment and long spin coherence. The T centre has an optical transition at telecom wavelengths, so each qubit can absorb or emit single photons that carry quantum information. The spins can be controlled with laser pulses and, as Photonic showed in 2025, with electrical signals through on-chip p-i-n diode structures. Photonic integrates these qubits into silicon photonic chips, where nanofabricated waveguides, gratings and cavities route photons between qubits and out to optical fibre. Each module is a silicon photonic chip with several T-centre qubits, an optical circuit and electronic controls. Qubits on the same chip are entangled through on-chip photonic circuits, and qubits on different chips or in different cryostats are entangled by sending photons through fibre. The result is a distributed quantum processor: a collection of smaller quantum nodes connected into one logical machine by optical entanglement. Photonic says its T centres keep their performance at about 1 kelvin.

Strengths – Silicon + Photons:

• CMOS Manufacturability: Because the qubits are built in silicon, Photonic can draw on decades of semiconductor process development. Implantation, lithography and CMOS processing can in principle produce qubit chips in large volumes with high precision, in contrast to superconducting circuits that need specialized fabrication or trapped ions that are assembled individually. If T-centre qubits can be reliably embedded across a wafer, a machine with thousands or millions of qubits could be assembled by tiling many silicon chips, work that existing microelectronics fabs already do. As researchers noted, “developing quantum technology using silicon provides opportunities to rapidly scale… since silicon photonics and electronics manufacturing is well-established”. Photonic’s Project VANGUARD proposal is an attempt to secure that manufacturing capacity in Canada, including advanced packaging.
• Native Telecom Networking: The telecom-wavelength optical interface, around 1326 nm in the O-band, is a major advantage. Photons at these wavelengths travel through standard optical fibre with little loss, which allows long-distance quantum links between qubits without the wavelength-conversion hardware that visible or near-infrared emitters need. In Photonic’s design, every qubit can send entangled photons directly into fibre. Modules can therefore be physically separate, even in different data centres, and still form one quantum computer, and distant users could share entangled qubits for quantum communication. The February 2026 teleportation over 30 km of TELUS’s installed fibre is the company’s first public result on a commercial network.
• High Connectivity for Error Correction: Photonic links let qubits within and across modules connect as a graph that is far richer than the 2D grid of many superconducting chips. Any two qubits that can exchange photons, directly or through intermediate nodes, can become entangled. Error-correction schemes such as QLDPC codes need exactly this kind of long-range connectivity, and Photonic’s hardware can in principle entangle non-neighbouring qubits through optical interference. The SHYPS simulations show what that could buy: the same logical performance with 2× to 3.5× fewer physical qubits at the code sizes tested, and much faster logical Clifford operations.
• Hybrid Control (Optical & Electrical): Photonic has demonstrated that its silicon qubits can be controlled both by light and by electrical pulses. The team showed the first electrically driven single-photon source in silicon, using a diode to excite a T centre and emit a photon, and used an electrically pulsed T centre to initialize a spin through the emitted photons. Wiring hundreds of qubits to electronic controllers, much as transistors are wired on a chip, scales far more easily than aligning hundreds of laser beams. Electrical control also allows many qubits to be triggered in parallel with synchronized voltage pulses, which becomes critical as qubit counts grow.
• Long-Lived Nuclear Spin Memory (Potential): Each T centre comes with nuclear spins, from its hydrogen atom and from nearby silicon-29 atoms, that can store quantum information for much longer than the electron spin. Researchers at UC Berkeley, Lawrence Berkeley National Laboratory and Dartmouth reported in an April 2025 preprint a three-qubit T-centre register in silicon photonics, with spin-echo coherence times of 0.41 ms for the electron, 112 ms for the hydrogen nuclear spin and 67 ms for a silicon nuclear spin. They entangled the two nuclear spins with a fidelity of 0.77, well short of error-correction grade. Long-lived memory helps a networked machine, because qubits must wait while entanglement between modules is set up. The Berkeley work is also one of several results from groups outside Photonic working on T centres.

Trade-offs and Challenges – “Networked” Quantum Computing:

• Complex Architecture: Photonic’s system is more complex than a single-chip quantum processor. It combines cryogenics, control electronics, photonic circuitry, defect qubits, fibre interconnects and synchronization of remote modules. Building each piece is hard; integrating them is harder. Fabricating chips with hundreds or thousands of T-centre qubits at specific locations, each coupled to a photonic cavity and a waveguide, pushes the limits of silicon defect engineering. Keeping qubit properties uniform (frequency, coherence, brightness) across a large chip or many chips is non-trivial. The complexity means more potential points of failure and more yield problems than a simpler monolithic device.
• Cryogenic Operation: The T-centre qubits still need cryogenic cooling. Photonic says they perform well at about 1 K, which avoids the dilution refrigerators superconducting qubits require, but a large machine would still need many cryostats and fibre connections between them. Photonic will need to show that its modules can be cooled reliably and run around the clock, and that fibre feedthroughs between cryostats do not add too much heat load or optical loss. The company says its systems can integrate into standard data centres; that has not yet been shown at scale.
• Entanglement Rate and Latency: In distributed quantum computing, the speed of two-qubit operations between modules is limited by how fast entanglement can be distributed. Optical entanglement between remote solid-state qubits has historically run at modest rates. Photonic’s public target is about 200 kHz at 99.8% fidelity, which would be a large improvement, but even 200 kHz is slower than on-chip gate speeds on some other platforms (which can reach MHz). Photonic can offset this by running many entanglement attempts in parallel over multiple fibre channels, and by using error correction that tolerates some latency. Reaching hundreds of kilohertz requires very efficient photon collection, low-loss optics and fast detectors, and every lost photon or detector dead-time slows the rate. Other groups are working on the same problem: IonQ linked two commercial trapped-ion systems with a photonic interconnect in April 2026, and QuTech researchers reported in January 2026 a teleported gate between remote diamond qubit registers that works unconditionally, on every run rather than only on selected ones.
• Probabilistic Entanglement & Overhead: The optical entanglement between qubits, created by interfering photons and heralding the result, is probabilistic: not every attempt succeeds. Photonic’s design has to retry failed attempts or run several in parallel, which adds overhead and complicates the scheduling of operations. Failed attempts are known because no detector clicks, so they mostly cost time. But if many operations in a computation depend on probabilistic entanglement, runtime becomes slow or variable. Photonic aims to raise success rates through better photon collection and coupling, and it needs fast classical control to manage distributed operations in real time.
• Defect Qubit Challenges: Each T centre is an atomic-scale defect inside the crystal, and variations in its local environment cause inhomogeneity, such as slight frequency differences or spectral diffusion (frequency drift) of the emitted photons. Photonic needs to tune and stabilize each qubit, through applied electric fields, strain or feedback, and has published work on electric-field tuning. Isotopically purified silicon-28 removes most nuclear-spin noise, but residual impurities or fabrication damage can still shorten coherence. High-fidelity two-qubit gates between spins, with error rates low enough for error correction, have not yet been demonstrated at scale; much of the published work so far concerns linking and initializing qubits.
• Competition and Technological Risk: Photonic’s approach is less proven than superconducting qubits or trapped ions, which already run circuits on 50 to 100 or more qubits. Unforeseen obstacles in materials, noise or integration could slow it down. Other silicon companies are ahead on delivered systems: Quantum Motion delivered a full-stack silicon CMOS quantum computer to the UK’s National Quantum Computing Centre in September 2025, and Diraq announced plans to install a quantum computer inside an Equinix data centre. In networking, IonQ is building photonic interconnects between trapped-ion systems, and Photonic is not among the performers in DARPA’s HARQ programme on heterogeneous quantum architectures. PsiQuantum is pursuing an all-photonic route, and IBM, Google and Quantinuum all publish dated fault-tolerance roadmaps. Photonic’s bold claims raise expectations, and any significant delay would invite scepticism.

In summary, Photonic’s silicon photonic modality offers scalability and connectivity advantages, but it requires uniting two frontier technologies, quantum computing and quantum networking. The strengths (manufacturability, telecom networking and inherent modularity) make it one of the more scalable architectures on paper. The trade-offs are the practical difficulties of implementation and of making a distributed system run as fast and as accurately as a monolithic one. If Photonic succeeds, the payoff is a path to scaling qubit numbers by adding more networked modules, an “infinitely scalable” quantum computer built like an internet of quantum nodes, in the company’s own words.

Track Record

Photonic has a record of announcing technical milestones roughly when it said it would, though several of its most important claims are still company-reported or based on simulation.

In 2023, Photonic said it would demonstrate key features of its distributed architecture within a year, and by mid-2024 it had posted the distributed entanglement and teleported-gate results. It said efficient error correction was on its near-term roadmap, and in early 2025 it released the SHYPS codes, which passed peer review in 2026 with smaller qubit savings than the launch figure. The underlying science has a longer publication record: T-centre characterization in PRX Quantum in 2020, single T centres in silicon photonics in Nature in 2022, the architecture in PRX Quantum, and electrical control in Nature Photonics in 2025. Researchers holding both SFU and Photonic affiliations have turned lab results into prototype devices quickly. Two gaps stand out in 2026: the 2024 distributed-computing paper has not appeared in a journal, and the TELUS fibre result has not been written up at all.

On the corporate side, Photonic’s funding has grown with its milestones, from about US$140 million by 2023 to more than US$350 million after the 2026 round, which valued the company at US$2 billion. In 2023, BCI’s CIO said Photonic had “reached several major technical milestones” and established key partnerships since BCI’s initial investment. Between November 2025 and May 2026, the company passed DARPA’s Stage A review, was selected for Canada’s Quantum Champions Program, and added Canada’s development bank, its export credit agency, Bell and TELUS to its investors.

Photonic has also built a large team. It has grown to more than 180 people across Canada, the US and the UK, and it announced a UK R&D facility in 2025. Its leadership is active in the quantum community; Dr. Simmons has co-chaired the advisory council for Canada’s National Quantum Strategy and speaks regularly at major conferences. The 2026 leadership changes brought in experience from outside quantum physics: a consumer-technology CEO, a cryptography entrepreneur and former national-security science adviser as chair, and executives from Microsoft Quantum and GCHQ for marketing and government affairs.

Photonic has not publicly missed a dated milestone, partly because it has published few dates. Its one dated public target, a distributed fault-tolerant solution within five years of 2023, falls due around 2028. Industry analysts have taken note of the company’s progress: in 2024, Global Quantum Intelligence’s David Shaw remarked that Photonic’s demos “set a new bar for quantum roadmaps that others will be under pressure to follow… [this] stands to accelerate the industry”.

Photonic tackled the hardest parts of the problem early, networking and error correction, and investors and governments have rewarded it with capital and scrutiny. Over the next two years it needs to turn that support into hardware results: a logical qubit, published link rates, and a system someone outside the company can use.

Challenges

Photonic still faces hard problems on the road to a commercial quantum computer:

• Scaling Up Qubit Production: Published experiments still involve a handful of T-centre qubits, and Photonic has not disclosed a physical-qubit count for any system. A useful machine will need chips containing perhaps thousands of T-centre qubits, all working reliably. Implanting or fabricating that many identical defect qubits in silicon with high yield is uncharted territory, and each qubit must couple to control electrodes and photonic structures. Photonic’s use of silicon and CMOS methods gives it a fighting chance, and Project VANGUARD shows it wants its own manufacturing capacity. Until a prototype with 50 to 100 qubits is demonstrated, scalability remains an open question.
• Maintaining Qubit Quality at Scale: Even if qubits can be made in quantity, their coherence and fidelity must stay high as the system grows. More qubits and components mean more sources of error, such as optical loss, crosstalk and spectral diffusion. The SHYPS simulations show how demanding this is: the larger SHYPS code beats a comparable surface code only below a physical error rate of 0.05%. Every beam splitter and fibre connector adds loss or noise. Photonic has to improve photon collection, stabilize qubit frequencies and use active feedback or error suppression, without trading quality for quantity.
• System Integration and Engineering: Photonic’s end goal looks more like a quantum data centre than a single box: many cryogenic modules, each with perhaps dozens of qubits, linked by fibre with precise timing and classical orchestration. Thermal management, vibrational stability for the optics, and timing synchronization for photon interference all need solving. The classical control architecture that manages operations across the network must be fast and coordinated, distributing clock signals and feed-forward information among nodes. Photonic has to build a distributed computing system in which the processors are quantum modules and the interconnect is optical, and a failure in synchronization or communication could decohere fragile quantum states.
• Achieving Fault-Tolerance Thresholds: Photonic’s QLDPC codes promise lower overhead, but the larger SHYPS code simulated so far has a pseudo-threshold of about 0.35%, lower than the comparable surface code, and it pays off only well below that. Photonic must bring physical error rates into the regime where logical qubits are stable and beneficial. Implementing QLDPC codes in hardware is a heavy lift in its own right: the published simulations use a belief-propagation and localized-statistics decoder that would have to run in real time, alongside many parallel operations and more complex syndrome measurements. Turning the SHYPS results into logical qubits on real hardware, with all error sources in play, is the central technical challenge.
• Timeline and Execution Risk: Photonic’s 2023 timeline, five years to a fault-tolerant solution, is now about two years from its due date, and no logical qubit has been demonstrated. Meeting it would require near-flawless execution. A material problem, a delayed component or difficulty hiring specialists could push the schedule out, and the quantum industry is full of projections that slipped. So far Photonic has delivered its early milestones, but the complexity grows sharply from here: routing thousands of fibres and calibrating thousands of qubits are problems nobody has solved for this platform. Because Photonic has promoted a faster timeline than most, a significant delay would affect how investors and customers see it.
• Commercial Proof: Photonic’s visible customers so far are governments (DARPA and Canada) and companies that are also its investors (Microsoft, TELUS, Bell and RBC). It has not named an enterprise customer outside that circle. The 2026 leadership changes and commercial hires show a go-to-market team being built, and the Microsoft resource-estimation work is its first offer aimed at developers and enterprises. Project VANGUARD, if funded, would add a large manufacturing business to run alongside the quantum roadmap, with the capital and management demands that brings.
• Competition and First-to-Market Pressure: Competitors are moving too. IBM, Google, Quantinuum and IonQ publish fault-tolerance roadmaps; Quantum Motion and Diraq are delivering silicon systems into national labs and data centres; PsiQuantum and Xanadu pursue photonic approaches; and IonQ and others are building photonic links between modules, conceptually close to Photonic’s distributed approach. A competitor could reach a large error-corrected system first and narrow Photonic’s window. Photonic will need to demonstrate tangible advantages, such as significantly lower qubit overhead or easier scaling, in hardware.
• Regulatory/Security Considerations: As Photonic’s technology matures, its dual-use nature (powering breakthroughs, but also breaking cryptography) will attract security scrutiny and possibly export controls. The company already works with DARPA and takes money from a Canadian defence-focused fund, which means parts of its work could become restricted. Its investor base also spans Five Eyes defence funders and an Abu Dhabi sovereign investor, a mix that allied defence customers may examine. Managing IP and government collaboration while commercializing broadly will be a non-technical challenge that shapes the company’s path to market.

In light of these challenges, Photonic’s journey is far from guaranteed to be smooth. As I previously noted, “there are many engineering challenges remaining… we’re not at a cryptographically relevant quantum computer yet, and this achievement doesn’t suddenly enable code-breaking – it demonstrates a new capability that could make future processors easier to scale”. That still describes Photonic’s situation in 2026: each advance is a stepping stone, and many more steps must follow. Photonic has identified the major hurdles, scalability, connectivity and error correction, and is working on them early. If it keeps gathering talent and support and starts publishing hardware results at scale, it could overcome them.

From the author