What if the hardest part of running a quantum computer isn’t the physics, but just plugging the thing in? ORCA Computing, a UK photonic quantum firm spun out of Oxford University, launched its newest machine, the ORCA PT-3, on October 15, 2026. The first unit is already live at a Digital Realty data center in London, sitting next to NVIDIA GPUs. That part is confirmed by multiple outlets covering the launch. What’s less settled is how big a deal the performance numbers really are, since they come from ORCA itself.
ORCA PT-3: A Quantum Computer That Plugs Into a Standard Rack
Most quantum computers need a dilution refrigerator, a vacuum chamber, or both. ORCA’s PT-3 needs neither. It encodes information in photons, particles of light, rather than in superconducting circuits or trapped ions. That lets it run at room temperature, and Data Center Dynamics reports the whole system weighs about 800 kilograms, draws roughly 6 kilowatts, and fits across two standard server racks.
The first PT-3 sits at Digital Realty’s LHR19 facility in London’s Docklands, right beside NVIDIA accelerated computing hardware. According to James Fletcher, ORCA’s head of solutions architecture, installation took about three hours, and the machine was fully running within 24 hours. No special flooring, power, or cooling retrofit was needed, the company says.
Co-founder and CEO Richard Murray explained why location matters so much for this setup:
“We’re leaning into the things that photonics does best in order to deliver performance,” Murray told Data Center Dynamics, adding that a hybrid AI model running quantum and classical hardware together “does not operate anywhere near as effectively over a remote link.”
That’s the real engineering bet here. Keep the quantum chip and the GPUs close enough that signals don’t have to travel far, and the hybrid system runs faster. Digital Realty’s Séamus Dunne called the deployment “a significant step towards commercial quantum advantage at scale,” though that phrase describes a goal ORCA has set for itself, not something it has proven yet.
Why It Matters to You
- Not a consumer product, not yet. This machine lives in an enterprise data center. There’s no app, subscription, or public access for ordinary people today.
- The indirect payoff is in things people rely on without seeing. ORCA says one PT-3 project is helping design new proteins, and another is working on scheduling power plants more efficiently. If either pans out, the long-run benefit could show up as cheaper medicine development or a more reliable, lower-cost electric grid, years down the line.
- Status check: both projects are described by ORCA as active pilots with named partners (NVIDIA on the protein work, JIJ Inc. on energy scheduling), not finished products anyone can buy or use.
Picture a hospital pharmacist named Dana, an illustrative example and not a real person, who wonders whether any of this quantum computing news ever touches her job. The honest answer right now is: not yet, and not directly. But if a system like PT-3 keeps helping researchers screen drug candidates faster, the lag between a promising lab result and a pharmacy shelf could eventually shrink. That’s a real possibility raised by this kind of work, not a promise anyone has kept.
What Got Faster, and By How Much
ORCA’s headline numbers compare the PT-3 to its own prior machine, not to a classical computer. On a 25,755-variable power-grid scheduling problem developed with bp and the UK’s National Quantum Computing Centre, ORCA says the hybrid workflow now finishes in about an hour, more than 12 times faster than on the previous-generation PT-2, at similar solution quality. Separately, a hybrid version of NVIDIA’s Proteína protein-design model, run across 32 NVIDIA Hopper GPUs paired with the PT-3, produced 20% more useful protein designs than a classical-only baseline, according to ORCA.
| Fact | PT-3 | Context |
|---|---|---|
| Weight | About 800 kg | Fits two standard server racks |
| Power draw | About 6 kW | Roughly 60% of one NVIDIA DGX H100 node’s draw, per Data Center Dynamics |
| Photonic capacity | 64 photons across 128 qumodes | Used in the protein-design project |
| Install to running | About 24 hours | Roughly 3 hours for physical installation, per ORCA’s James Fletcher |
Source: ORCA Computing announcement, reported by The Quantum Insider and Data Center Dynamics, Oct. 15, 2026.
This isn’t the first time a quantum company has tried pairing its hardware tightly with NVIDIA’s. Alice & Bob recently tested decoding its own error-correction codes on NVIDIA GPUs, chasing the same idea that fast classical hardware sitting right next to a quantum chip unlocks more than either piece alone. ORCA’s approach is different, wiring an entire quantum processor permanently into a commercial data center rather than connecting it over a network link, the way IonQ’s photonic interconnects aim to join separate quantum chips together.
What We Don’t Know Yet
- Every performance figure here, the 12x speedup, the 20% protein-design gain, the “no modification needed” installation claim, comes from ORCA. None of the outlets covering the launch, including The Quantum Insider and Quantum Computing Report, report independent testing of these numbers.
- The 12x speedup is measured against ORCA’s own older machine, not against a classical-only supercomputer. Whether PT-3 actually beats a top classical solver on this problem, the harder and more meaningful bar, isn’t addressed in the coverage.
- The 20% protein-design figure doesn’t come with a clear definition of “useful designability,” so it’s hard to judge how big that gain really is in practical terms.
- ORCA’s press materials use the phrase “commercial quantum advantage.” That’s a stated goal, not a claim that’s been independently verified or peer-reviewed.
- No price has been published for a PT-3 system, and ORCA hasn’t said how many more customers plan to deploy one.
A photonic quantum computer that behaves like ordinary data-center gear is still a meaningful engineering step, even if the performance claims need outside scrutiny. Getting hardware out of a physics lab and into a commercial rack, without a cryostat crew on call, is exactly the kind of boring-sounding progress that tends to matter more than it sounds like it should.
Worth watching next: whether any of ORCA’s pilot customers publish their own results independently of the company, and whether a rival photonic or trapped-ion system shows up in a data center on similarly short notice. If colocation turns out to be the real unlock for hybrid quantum-AI work, expect more of these announcements, fast.
Sources
- The Quantum Insider: ORCA Computing Launches PT-3, The Latest Generation of Its Photonic Quantum Computers (Oct. 15, 2026)
- Data Center Dynamics: Photonics Quantum Firm Orca Computing Deploys PT-3 System at Digital Realty’s LHR19 Data Center (Oct. 2026)
- Quantum Computing Report: ORCA Computing Launches PT-3 Photonic Quantum System in London for Hybrid AI and Industrial Optimization (Oct. 2026)
