Summary

• The operational PT-3 system runs alongside NVIDIA computing infrastructure at Digital Realty's London Innovation Lab.
• ORCA reports a 20% improvement in a specified protein-design metric against a classical model baseline.
• Its energy optimisation workflow ran more than twelve times faster than on PT-2 at equivalent solution quality, according to the company.

British technology company ORCA Computing has installed its PT-3 photonic quantum computer in a commercial London data centre, combining specialist quantum processing with NVIDIA accelerators to test applications in scientific modelling and industrial optimisation.

The system is operational at Digital Realty’s Innovation Lab in London Docklands, where the company is developing hybrid applications that use quantum and conventional computing together. ORCA announced the installation on 15 October, alongside results from protein-design and electricity scheduling benchmarks.

PT-3 uses photonic technology, processing quantum information through states associated with light. It operates at room temperature and occupies two standard data centre racks, with ORCA specifying approximately 6kW of system power consumption.

Those figures relate to the PT-3 equipment and should not be interpreted as the electricity demand of the accompanying NVIDIA infrastructure or the wider installation. The company’s architecture is intended to make quantum processors easier to integrate with conventional computing environments.

ORCA describes PT-3 as designed for commercial quantum advantage, but that remains the company’s objective and positioning rather than a generally demonstrated result. Its published experiments provide evidence of specific hybrid benchmarks, not proof that the machine outperforms every relevant classical alternative.

Hybrid computing is central to the proposition. Conventional graphics processors continue to perform significant parts of the workload, while the quantum system is used for operations that may provide an additional computational capability.

Protein design provides the first benchmark

ORCA extended NVIDIA’s Proteína generative AI model with a photonic quantum component, creating a hybrid system intended to explore protein structures. The reported experiment involved 32 NVIDIA Hopper GPUs and a PT-3 configuration using 64 photons across 128 quantum modes.

The company says the combined system improved the proportion of useful protein designs by 20% against its specified classical baseline. It also reported a different distribution of generated secondary structures, suggesting that the hybrid model explored a different range of candidates.

The performance figure concerns a computational designability metric rather than a demonstrated improvement in drug effectiveness, laboratory discovery rates or commercial research costs.

Predicting potentially useful protein structures is an early stage in a much longer scientific process. Computational candidates generally require further assessment and, where appropriate, laboratory validation before researchers can establish whether they exhibit the intended physical or biological properties.

The comparison also involves a hybrid architecture. Since NVIDIA GPUs form part of the tested system, the results cannot be presented as evidence that PT-3 acting alone outperformed a conventional computer.

ORCA has previously reported research using PT-2, including studies involving peptide binding and molecular generation. Those experiments provide technical background, but they used different methods and benchmarks and should not be treated as direct proof of the performance of the new system.

The relevance of PT-3 will depend on repeatability and comparison with well-chosen classical alternatives. A technique can improve one model or dataset without delivering the same advantage across other scientific workloads.

Energy scheduling tests a different requirement

The second application involves unit commitment, an optimisation problem used in electricity generation planning. Operators need to determine which generating resources should run over particular periods while meeting demand and respecting operational constraints.

Such calculations can become computationally demanding because different decisions interact. The choice to operate one generator may influence costs, equipment limits and the availability of other resources later in the scheduling period.

ORCA worked with Japanese optimisation company JIJ on a hybrid approach, building on earlier research involving bp and the UK’s National Quantum Computing Centre.

The latest benchmark concerns a problem with 25,755 variables. ORCA says PT-3 completed its hybrid workflow in approximately one hour, more than twelve times faster than its previous-generation PT-2 system while maintaining the same solution quality.

That is a comparison between two generations of ORCA quantum hardware. It is not a finding that PT-3 was twelve times faster than conventional optimisation software.

The company separately says the hybrid approach achieved solution quality competitive with a leading commercial optimisation solver. Earlier research also examined what happened when different methods were allowed additional runtime, but those results do not establish a universal speed or cost advantage for the new installation.

Commercial electricity operations would impose further requirements. A production system must work with forecasts, network conditions, asset constraints and procedures designed to preserve reliability. Performance on a defined benchmark is a useful development result, but it is not evidence that a live grid is being controlled by the technology.

Quantum computing enters conventional infrastructure

Locating PT-3 in Digital Realty’s facility gives researchers and prospective customers an environment where quantum equipment can be used alongside existing high-performance computing resources.

The installation avoids some practical barriers associated with operating specialist research hardware in separate facilities. However, customers must still determine which parts of their applications are suited to quantum processing and how the additional equipment affects data movement, software integration and total operating costs.

ORCA has identified customer trials in sectors including life sciences, energy, defence and logistics. Those activities indicate potential commercial applications, but trials must be distinguished from independently verified production deployments with demonstrated economic returns.

The ability to upgrade the hardware in the field forms another part of ORCA’s strategy. It could help customers expand capabilities without replacing an entire installation, although the effect on performance and cost will depend on future upgrades and the applications they support.

PT-3 is an installed and operating system, and ORCA has published defined results from two different technical experiments. The next test is whether repeated comparisons against appropriate conventional methods establish practical advantages that justify adding photonic quantum processing to commercial computing workloads.