EY Installed a Quantum Computer in Toronto. Nobody Will Say What It Is.

July 29, 2026 – On July 29, 2026, EY announced it had installed an on-site quantum computer at its Toronto office, led by EY Canada, as part of a global investment of more than USD 3 billion in AI and “next frontier technologies.” The EY Canada release described the new capability as supporting “the processing of highly sensitive workloads in areas such as optimization, fraud detection, data protection and large-scale risk management.” Joe Depa, EY’s Global Chief Innovation Officer, told Accounting Today that, to his knowledge, EY is the first Big Four firm to own a physical quantum computer rather than relying on cloud access or vendor partnerships.

The announcement named no vendor. It disclosed no qubit count, no gate fidelity, no modality, no technical specification of any kind. I reached out to both EY’s media contact and CTO Biren Agnihotri requesting vendor, model, and architecture details. I received no response. Much of the initial coverage repeated EY’s release without adding technical detail.

Accounting Today’s Chris Gaetano obtained one important additional fact from Depa: the system is a photonic quantum computer. Depa declined to share exact specifications or cost. He said EY chose Canada because of its recognized leadership in quantum computing, access to talent, and proximity to clients. He also stated that the immediate focus is on post-quantum cryptography readiness, not on running production workloads. As Depa put it, the purchase was “made with more of an eye toward the future than the present.”

That last sentence is the honest framing. The press release tells a different story. The gap between the two is worth examining.

Two Plausible Vendors, Zero Confirmation

Depa confirmed “photonic.” The machine is in Toronto. Who could have supplied it?

Xanadu Quantum Technologies is the most obvious candidate. Xanadu is a Canadian photonic quantum computing company headquartered in Toronto that went public on Nasdaq and the TSX on March 27, 2026, under the ticker XNDU. It builds photonic quantum computers using squeezed light and continuous-variable encoding, and has two systems (Borealis and Aurora, discussed below). Xanadu’s 2026 strategic messaging has centered on Canadian data sovereignty: in March 2026, it signed an MOU with TELUS to explore sovereign quantum data center infrastructure, using language about keeping “critical data and intellectual property in Canada, under Canadian control” that echoes EY’s own data-residency framing. Xanadu was selected for Canada’s Quantum Champions Program (up to CAD 23 million), advanced to Stage B of DARPA’s Quantum Benchmarking Initiative, and reports Q2 2026 earnings on August 5.

But there is a second candidate that several commentators have overlooked.

ORCA Computing is a UK-headquartered photonic quantum computing company with an office in Toronto. ORCA builds room-temperature photonic systems using time-bin-encoded single photons, a different photonic architecture from Xanadu’s squeezed-light approach. Critically, ORCA has an existing working relationship with EY. ORCA’s own applications page lists EY as having “applied generative modelling techniques on the PT Series for onshore and offshore wind farm location optimisation.” EY UK separately described “a research project with Innovate UK and ORCA Computing to assess the feasibility of a hybrid quantum and AI solution to predict wind patterns for renewable energy production.” ORCA had delivered 10 on-premises PT-Series systems to customers worldwide by early 2025, making it the photonic vendor with the most commercial on-premises deployment experience.

Both vendors are photonic. Both have a Toronto presence. ORCA has a documented working relationship with EY and a track record of commercial on-premises deployments. Xanadu is headquartered in Toronto with a stronger Canadian-sovereignty narrative. Neither has been confirmed as the vendor. The honest conclusion is that I do not know which company supplied EY’s machine, and neither does anyone outside EY and the vendor.

That uncertainty is itself the story. EY has placed data location, quantum experimentation, and computational utility into the same narrative without disclosing the technical facts needed to evaluate any of them.

What Each Vendor Could Have Provided

The identity of the vendor determines what EY has, because the two companies build very different machines.

If the system is from Xanadu, EY likely has some variant of one of two platforms.

Xanadu’s Borealis uses 216 squeezed optical modes and demonstrated quantum computational advantage in a 2022 Nature paper, solving a Gaussian boson sampling problem that Xanadu estimated would take a contemporary classical supercomputer roughly 9,000 years to simulate by the specific classical method they analyzed. Borealis was previously available on Xanadu Cloud and Amazon Braket, though it is no longer publicly accessible through those platforms. It is a special-purpose sampling device, not a universal gate-based computer. No public result shows it delivering classically advantaged performance on any business workload.

Xanadu’s Aurora, published in Nature in January 2025, is a modular photonic architecture consisting of four server racks, 35 photonic chips, and 13 km of fiber optics. It produces 12 physical qubit modes per clock cycle and demonstrated the subsystems needed for fault-tolerant photonic computation, including a distance-2 repetition-code experiment with real-time decoding and feedforward. (Xanadu separately demonstrated 12 GKP-encoded logical qubits with real-time error correction in a different 2025 Nature paper, which is a distinct result from the Aurora architecture itself.) Aurora runs at room temperature, though high-performance photon-counting detectors require cryogenic cooling. Its modular design, where additional racks connect via standard fiber optics, is photonics’ strongest scaling argument. Xanadu claims the approach can scale to “thousands of server racks and millions of qubits.”

Aurora matters architecturally. But 12 physical qubit modes, or even 12 GKP-encoded logical qubits, cannot perform optimization, fraud detection, or risk management at any commercially relevant scale. Xanadu’s roadmap targets up to 500 logical qubits by 2029-2030. Its 2025 annual revenue was USD 4.6 million.

If the system is from ORCA, EY likely has a PT-2 or the newer PT-3 system. ORCA’s PT-Series machines are rack-mounted, room-temperature photonic processors that have been deployed at the UK’s National Quantum Computing Centre, Montana State University, and enterprise customers in Japan (through Toyota Tsusho). ORCA has demonstrated practical applications in network optimization (Vodafone) and cybersecurity (ST Engineering partnership). The PT-3 targets quantum advantage on optimization and generative AI workloads. ORCA’s commercial model is built around on-premises deployment, with installations completed in under two days.

In either case, the machine in EY’s Toronto office is a research-grade photonic system at the earliest stages of the technology’s development. Neither vendor’s current hardware can deliver production-scale, classically competitive results on the workloads named in EY’s press release.

The CRQC Scorecard Perspective

For readers who follow my work on quantum security: under the definitions used in my CRQC Scorecard, the photonic modality has the largest gap to a cryptographically relevant quantum computer (CRQC) of any modality on every metric I track. No photonic system has demonstrated a logical qubit that meets the criteria in the CRQC Quantum Capability Framework. The Logical Qubit Capacity, Logical Operations Budget, and Quantum Operations Throughput gaps are all effectively the entire journey.

For comparison: Quantinuum has demonstrated 94 error-detected logical qubits on a trapped-ion system (using high-rate iceberg codes with distance-2 error detection). The Harvard/MIT/QuEra collaboration demonstrated up to 96 logical qubits on a neutral-atom platform using a high-rate [[16,6,4]] code. Google’s Willow chip achieved a surface-code cycle time of 1.1 microseconds, a figure that matches the throughput target in my CRQC framework. QuiX Quantum’s April 2026 demonstration of below-threshold photon distillation put photonics on the board for error suppression, but that result operates at the pre-QEC physical layer, not at the logical-qubit level.

My assessment places a photonic CRQC at 15+ years away under current trajectories. I also noted that photonics is the modality most likely to produce a surprise that changes the assumptions underlying other modalities’ roadmaps. Both of those judgments still hold, and neither changes the fact that what EY has in its Toronto office is, by any honest technical measure, an early-stage research instrument.

The Press Release vs. The Interview

I want to be precise about what I am criticizing and what I am not.

Depa’s comments to Accounting Today were measured and honest. He said the focus is on “building capability and readiness.” He acknowledged that “the immediate value isn’t replacing classical computing.” He framed the purchase around future positioning, talent development, and post-quantum cryptography. That framing is legitimate, and it aligns with a pattern I analyzed at length in Why Companies Buy Quantum Computers That Can’t Do Anything Yet.

The press releases are different documents. They claim the machine will support “the processing of highly sensitive workloads in areas such as optimization, fraud detection, data protection and large-scale risk management.” The EY Canada version goes further, describing “a shift from quantum experimentation to solving real business challenges for clients.” EY has provided no technical evidence that the installed system can deliver production-scale, classically competitive results in any of these areas. No benchmark. No classical comparison. No vendor. No specifications.

This is the kind of gap between careful private framing and expansive public messaging that erodes trust in the quantum industry. A CISO reading the press release would reasonably conclude that EY has acquired a machine performing commercially relevant computations on sensitive data. A CISO reading Depa’s interview would understand that EY has bought a learning platform. Those are different stories.

Why It Might Still Be a Good Idea

As I argued in my analysis of why companies buy quantum computers that can’t do anything yet, there are rational reasons for this purchase.

Procurement readiness. In large organizations, the path from evaluating a technology to operating it takes years. EY is ensuring that when the next generation of photonic hardware arrives, its facilities, security accreditation, integration patterns, and vendor relationships are already in place.

Talent. Quantum talent is scarce. Owning real hardware attracts researchers and engineers. The institutional learning curve for quantum computing cannot be accelerated by reading papers or running simulators. Real hardware has noise, calibration drift, device-specific behavior, and operational complexities that only hands-on experience teaches. (Cloud access to physical QPUs can teach some of this, but on-premises ownership adds facilities management, maintenance, physical security, control-plane integration, and operational governance.)

PQC readiness. Depa cited post-quantum cryptography as the immediate focus. For a Big Four firm whose clients include banks, governments, and regulated enterprises working toward NIST’s finalized PQC standards and NSA’s CNSA 2.0 requirements for National Security Systems, building in-house quantum expertise is a legitimate business investment. PQC migration principally involves cryptographic discovery, protocol analysis, crypto-agility engineering, and governance, all of which run on classical infrastructure. Owning a photonic system does not directly accelerate those activities. But it may contribute to broader quantum literacy, client engagement, and a more informed view of the technology’s trajectory.

Quantum systems integration and security. This is a less obvious rationale, but to me the most interesting one. Running on-premises quantum hardware teaches an organization about control planes, software supply chains, identity and workload access, physical security, remote maintenance, telemetry, vendor dependencies, incident response, and the operational security of quantum infrastructure itself. Those lessons are learnable now and cannot be reduced to PQC alone.

Canadian data residency. An on-premises system can reduce some data-residency concerns for EY Canada’s regulated-industry clients, depending on its control plane, telemetry, support model, and external dependencies. EY has not disclosed enough architectural information to determine whether the deployment is operationally self-contained.

Each of these reasons is valid on its own terms. The purchase may well be a rational capability-building investment. It is not what the press release describes.

What to Watch

August 5: Xanadu’s Q2 2026 earnings. If EY’s system is from Xanadu, the deal may appear in Xanadu’s quarterly results or earnings call. For a company with USD 4.6 million in 2025 annual revenue (though Q1 2026 revenue already jumped to CAD 2.8 million, roughly 4x year-over-year), even a moderately sized hardware sale would be material.

ORCA’s deployment announcements. ORCA has been steadily disclosing new customer deployments (Toyota Tsusho in Japan, NQCC in the UK, Montana State in the US). If EY is an ORCA customer, ORCA’s communications cadence suggests a public announcement would follow.

EY Canada follow-up disclosures. EY publicized a patent in February 2026 for hybrid classical-quantum methods (originally filed April 2023, US-12488263-B2), with examples referencing quantum annealing and QUBO-type methods. Whether subsequent technical publications or patent filings reference photonic hardware specifically would help confirm the platform.

Competing Big Four moves. EY’s announcement creates pressure on Deloitte, PwC, and KPMG to demonstrate their own quantum capabilities. All four firms have quantum partnerships and cloud access arrangements across multiple platforms. Whether competitors respond with their own on-premise deployments or dismiss EY’s move will signal how the professional services market views the near-term quantum value proposition.

The Bottom Line

EY installed a photonic quantum computer in Toronto. The vendor, model, architecture, qubit count, fidelity, programming model, and benchmark performance remain undisclosed. Two photonic vendors, Xanadu and ORCA Computing, are plausible suppliers based on the public record. Neither has been confirmed.

Whatever the vendor, no current photonic quantum computer can deliver production-scale or classically advantaged performance in optimization, fraud detection, data protection, or risk management. The system should be understood as an early-stage photonic platform for experimentation and operational learning, not as the enterprise computing capability implied by EY’s press release.

That is still a defensible reason to acquire one. An on-site system can help EY’s teams learn about quantum hardware integration, build operational expertise, develop a more informed view of an emerging technology, and strengthen its PQC and quantum-readiness advisory practice. Those are commercially rational investments in an organization that advises Fortune 500 clients on technology risk.

The problem is not that EY bought a quantum computer. The problem is the distance between what Depa told a journalist and what EY told the world.