Research Notes

Is Diraq’s Data Center Quantum Computer a Product or a Field Test?

Research Finder

Find by Keyword

Is Diraq's Data Center Quantum Computer a Product or a Field Test?

Eight qubits, under 20 kW, no fidelity data from the floor. Why the measurement campaign matters more than the headline.

11/18/2026

Key Highlights

  • Diraq has installed an eight-qubit silicon spin quantum computer inside an Equinix data center in Sydney. The cryogenics and control electronics are self-contained, and the system draws under 20 kW across roughly 172 square feet.
  • Diraq plans to swap in a processor of up to 256 qubits in the same system within the next year. It targets utility-scale machines in about five years, at no more than tens of millions of dollars each.
  • No customers, access model, pricing or on-site fidelity figures were disclosed. The stated purpose is to evaluate operation in a live facility, with remote monitoring and integration with classical CPUs and GPUs.
  • Diraq is one of six publicly identified companies in Stage C of DARPA's Quantum Benchmarking Initiative, alongside IBM and IonQ. It also holds a preliminary US$38 million CHIPS Act letter of intent.
  • We read the deployment as an environmental qualification exercise for spin qubits. Its value will be measured by published uptime and drift data, not by qubit count.

The News

Diraq has installed an eight-qubit silicon spin quantum computer at Equinix in Sydney, which it describes as the first of its kind in a commercial data center. The companies position the system as a test of whether quantum hardware can run alongside ordinary servers in shared space, with remote monitoring and links to classical CPUs and GPUs. Diraq says the installation is designed to accept a processor of up to 256 qubits within a year, a path it argues could lead to utility-scale systems in roughly five years. The full announcement is available on Diraq's newsroom.

Analyst Take

Quantum vendors have spent a decade promising that their machines will one day sit in ordinary data centers. Diraq has now put one there, and the honest scale of it is modest: eight qubits, no named users, and no fidelity figures from the facility floor. Skeptics who judge the install by what it can compute have a fair point. An eight-qubit computer can be simulated on the laptop of whoever drove over to look at it, so nothing calculated in that cage sits beyond classical reach, and a famous building lends a credibility the hardware has not yet earned. Computation, though, was never the thing under test. Spin qubits are tiny and sensitive to charge noise and temperature swings, and the modality's commercial case rests on packing them densely and running them almost anywhere. A lab cannot prove that. A live colocation floor, with cooling plant cycling and other tenants working around the cage, can.

What Was Announced

The system pairs a Diraq silicon chip carrying eight spin qubits with a self-contained dilution refrigerator and control electronics. Diraq builds its qubits as quantum dots using standard CMOS processes. In 2025, Diraq and imec reported in Nature that four Diraq-designed devices from imec's 300mm line exceeded 99% fidelity for one- and two-qubit operations, with state preparation and measurement reaching up to 99.9%. That went a long way toward answering whether foundry-made devices could match hand-tuned academic ones. It says nothing about the Sydney system, and Diraq has not published figures for it.

Nor is Diraq the first to get spin hardware out of the lab. Quantum Motion's full-stack silicon CMOS system at the UK's National Quantum Computing Centre, built on a tileable four-qubit unit cell, has been running under a government testbed program since September 2025. Diraq's "first" holds only because of the commercial data center qualifier. What changes in Sydney is the landlord: a commercial operator with paying tenants rather than a national facility built for experiments.

Power and floor space tell us more. Under 20 kW is roughly a sixth of the nominal draw of a single NVIDIA GB200 NVL72 rack, and about 172 square feet is closer to a small cage than a dedicated quantum hall. Diraq's upgrade path is a processor swap: up to 256 qubits in the same refrigerator within a year. That would be a 32-fold increase with minimal change to power and space, if it works as designed. That swap is the claim this deployment appears built to test. The longer-range figures, a utility-scale system near 300 kW and a cost ceiling in the tens of millions of dollars, are targets rather than engineering results. We are not aware of any Uptime Institute tier rating written with millikelvin equipment in mind.

Read the release closely and the use case shrinks. The companies list live-facility evaluation, remote monitoring, secure links to classical CPUs and GPUs, and readiness for hybrid workloads. The release pairs this with Diraq's own Quantum at Work paper on drug discovery economics, which models a saving of about US$600 million per approved medicine. Those use cases belong to machines several orders of magnitude larger. For now, the workload is the facility itself.

Market Analysis

We limit the comparison here to IBM and IonQ, the two Stage C peers with commercial systems in market, rather than attempting a full vendor landscape. IBM's Nighthawk carries 120 qubits on 218 tunable couplers, and IBM says the r2 revision supports circuits of more than 7,500 operations. IBM says it plans to extend that architecture beyond 1,000 qubits in 2027 and expects its fault-tolerant Starling system by 2029, with advantage claims submitted to an open community tracker. Those are roadmap items, and they deserve the same discount we apply to Diraq's. IonQ reported Q3 revenue of $173.8 million, nearly half of it from its SkyWater foundry segment. It raised full-year guidance to $505 to $515 million and carries $1.2 billion in remaining performance obligations. Its 99.99% two-qubit gate result, reported in October 2025 on R&D prototypes, sets the fidelity bar Diraq will be judged against.

On the hybrid story, IonQ sets the bar. Its Superion 256 is slated to be the first on-premise system at NVIDIA's Accelerated Quantum Research Center. That makes the quantum-GPU integration claims in Diraq's release, which names no specific accelerator stack, look general by comparison.

The CMOS manufacturing pitch no longer belongs to Diraq alone. IBM closed its acquisition of HRL Laboratories in August, which gives it a silicon spin program of its own alongside its superconducting roadmap. IBM has flagged potential plans to develop spin qubit manufacturing at Anderon, the 300mm quantum wafer foundry it is standing up in Albany, which starts with superconducting wafers. That is a direct counter-position to Diraq: the largest name in the field now holds spin expertise in-house and, prospectively, a fab to scale it, while IonQ, through SkyWater, already owns one. CMOS compatibility is heading toward table stakes, and the open question becomes who shows working density at scale first.

Diraq's strongest credential is DARPA's Quantum Benchmarking Initiative. It entered Stage C on 7 October with IBM, IonQ and Atom Computing, joining Microsoft and PsiQuantum, which arrived earlier through DARPA's US2QC pilot. Stage C moves from paper plans to independent hardware verification, and field data from a live facility is the kind of evidence that process is likely to value.

Equinix collects optionality cheaply. It has hosted Oxford Quantum Circuits' superconducting system at TY11 in Tokyo since 2023, so what is new in Sydney is the qubit type, not the venue. Like a crocodile at a river crossing, Equinix does not need to pick the best herd, only to be where every herd has to cross. The crossing is not free. Equinix has published no commercial terms for quantum colocation, has not shown tenant demand, and carries the operational burden of cryogenic plant in shared halls. Incentives pull in different directions as well. Equinix needs something it can sell, Diraq needs evidence for DARPA reviewers, and enterprise buyers want service levels no quantum vendor offers yet.

Looking Ahead

The key trend we'll be monitoring is whether Diraq publishes field data from Sydney. Uptime and fidelity drift tied to facility events would do more for the silicon spin thesis than any qubit count. They go straight to the deployability claim that separates the modality from superconducting and trapped-ion rivals. The 256-qubit processor swap promised within twelve months is the next test of that claim. Should it land inside the same refrigerator and power envelope with fidelities near the 2025 Nature results, Diraq would hold a data point its larger rivals have not yet published for a colocation setting. Timing helps, with Stage C verification underway and likely to weigh hardware evidence over roadmaps. Whether Equinix turns these placements into a priced quantum colocation service or keeps them as showrooms remains open. For CIOs there is nothing to procure yet, but there are questions worth putting to both companies now. What uptime did the system hold over its first 90 days? How far did gate fidelity drift through cooling cycles, power transfers, and maintenance windows? How often did it need recalibration with no engineer on site?

Author Information

Stephen Sopko | Analyst-in-Residence – Semiconductors & Deep Tech

Stephen Sopko is an Analyst-in-Residence specializing in semiconductors and the deep technologies powering today’s innovation ecosystem. With decades of executive experience spanning Fortune 100, government, and startups, he provides actionable insights by connecting market trends and cutting-edge technologies to business outcomes.

Stephen’s expertise in analyzing the entire buyer’s journey, from technology acquisition to implementation, was refined during his tenure as co-founder and COO of Palisade Compliance, where he helped Fortune 500 clients optimize technology investments. His ability to identify opportunities at the intersection of semiconductors, emerging technologies, and enterprise needs makes him a sought-after advisor to stakeholders navigating complex decisions.