ORCA Computing has switched on the PT-3, its newest quantum computer, in a London datacentre. The machine is a test deployment at the Digital Realty Innovation Lab that runs alongside Nvidia-accelerated computing. ORCA claims it is the first quantum system built to deliver what the company calls “commercial quantum advantage”.
The PT-3 runs at room temperature and draws 6 kW across two standard server racks, has a claimed install time of hours, and upgrades in the field. That sets it apart from superconducting machines from rivals such as IBM, Google and Finland’s IQM, which need cooling to near absolute zero inside cryogenic fridges.
Quantum computers handle information differently from the machines that run today’s IT. Where a classical computer works through a maze one path at a time – albeit very quickly – a quantum machine behaves more like water poured into the whole maze at once, weighing up many possible answers in the same moment. The PT-3 is photonic, which means it encodes information in light rather than superconducting circuits.
The company’s headline result is in drug discovery. ORCA extended Nvidia’s Proteína model, an AI system that designs proteins, the molecules behind most medicines. Trained on 32 Nvidia Hopper GPUs alongside the PT-3, the hybrid model produced 20% more protein candidates that passed real-world quality checks than the classical version, and it generated a broader variety of protein shapes.
The PT-3 has also tackled power scheduling, the question of which generators to switch on and when to meet demand at lowest cost. On a benchmark with 25,755 variables, co-authored with Japanese software firm JIJ, energy company BP and the UK’s National Quantum Computing Centre, the machine found answers in around an hour. This was more than 12 times faster than its predecessor at quality comparable with commercial mathematical optimisation software.
Earlier results point the same way. ORCA’s previous PT-2 system, tested with the Technical University of Denmark, produced around twice as many strong peptide binders as classical AI, a result confirmed physically in the lab, while the company also trialled network optimisation for telecoms providers. It generated 49% more valid molecules on the QM9 chemistry benchmark.
Richard Murray, co-founder and chief executive of ORCA Computing, said: “By integrating our photonic quantum computer with Nvidia-accelerated computing, we have taken a critical first step in the future of hybrid quantum supercomputing. The PT-3 is already delivering results in AI for science and industrial optimisation, from protein design to energy scheduling.”
The launch rests on years of public backing. The UK’s National Quantum Strategy has made quantum a national priority, with the government recently committing £1bn over the next four years to win the race to deliver quantum computing at scale. ORCA, spun out of the University of Oxford in 2019, claimed it was the first to place a quantum computer inside a datacentre in 2022, and it has now installed 11 PT Series systems in customer datacentres worldwide, working with more than 35 customers and partners, with banks such as NatWest testing the technology for fraud detection.
ORCA’s claims still need careful handling. The industry has not yet built a universal, fault-tolerant quantum computer, and that milestone remains a target for the late 2020s or early 2030s. Today’s machines sit in a phase analysts describe as “near-term hybrid advantage”, in which they act as specialist accelerators alongside GPUs and CPUs rather than replacing them.
Its benchmarks are supplier-run or whitepaper results, and are competitive with rather than clearly beating classical mathematical solvers, while the 20% figure rests on a single test.
