Why true scalability means 'a box, not a building'
Data centre power consumption in the US is forecast to reach 106 gigawatts by 2035, a figure that was revised upward by 36% in the space of seven months. Grid operators, investors and policymakers are all asking how computing infrastructure can keep growing without overwhelming energy systems that are already close to their limits.
Quantum computing is about to join that conversation. Most roadmaps to a million qubits describe the physics in detail but say very little about how large the resulting machine will actually be, or how much power it will draw. As quantum computers move from laboratory demonstrations toward commercially useful machines, the same scrutiny now applied to AI infrastructure will be applied to them, and the answers turn out to differ enormously depending on which hardware approach is being built.
For quantum computing to be commercially useful, it has to be something that can actually be deployed. That means fitting inside standard data centres, operating within realistic power and cooling budgets, and sitting alongside the HPC and AI infrastructure. Silicon spin qubits are the approach best positioned to meet that challenge, for reasons that come down to qubit density, operating temperature, and how much of the control hardware can be integrated directly alongside the processor.
The facility bottleneck: How much power does a quantum computer need?
4,000 logical qubits is often cited as the threshold for solving commercially useful problems. An independent economic analysis comparing power demand across qubit modalities found that superconducting systems would need around 160 megawatts, ion-trap systems around 140 megawatts and photonic systems around 100 megawatts.
For context, a modern hyperscale AI data centre runs at 100 to 200 megawatts. The demands created by quantum computers aren't incremental additions to a shared facility, they will need power plants built to serve a single machine. Some approaches in development call for sites spanning tens of thousands of square metres, housing thousands of modules with their own dedicated cryogenic plant and substation. That isn't a quantum computer that can share space with existing HPC and AI workloads. It is a data centre in its own right, built around one system.
Quantum Motion's target utility-scale system is designed to deliver that same 4,000 logical qubits inside five standard server racks, drawing under 200 kilowatts. That's roughly a thousandfold reduction against superconducting approaches at equivalent computational scale, and it sits comfortably inside the power demand of an ordinary enterprise data centre.
The 'Transistor Moment' On The Data Centre Floor
The gap comes down to three factors that compound on each other. Silicon spin qubits are small enough that millions can be patterned onto a chip the size of a fingernail. They operate at temperatures where cryogenic cooling is a manageable part of the power budget rather than the dominant one. And because the control electronics can be built using the same cryo-CMOS processes as the qubits themselves.
Image: silicon QPU chip shown on a fingertip for scale
Manufacturing qubits and their control circuitry on standard 300mm CMOS wafers, the same lines that produce smartphone and laptop chips, means quantum hardware can draw on fifty years of semiconductor manufacturing maturity rather than requiring bespoke fabrication, custom cryoplants or room sized laser arrays built from scratch for every new system. This is quantum computing's transistor moment.
The first silicon CMOS quantum computer in a data centre
In September 2025, Quantum Motion delivered the industry's first full-stack quantum computer built on 300mm silicon CMOS wafer technology to the UK's National Quantum Computing Centre at Harwell. The system, including its cooling infrastructure, cryogenic control electronics and room-temperature support hardware, fits inside three standard 19-inch server racks. The quantum processing unit at its centre is millimetres across.
Video: Quantum Motion Delivers the Industry's First Full-Stack Silicon CMOS Quantum Computer
Standard 19-inch racks are the building block of data centres worldwide. Quantum Motion has shown that a quantum computer in this form factor can be installed, powered, cooled and maintained through the same facilities processes that already support classical HPC and AI infrastructure, without purpose-built rooms or a dedicated cooling plant.
Fitting alongside HPC and AI
The most promising near-term uses of quantum computing are hybrid ones, where a quantum processor handles the specific calculation that's intractable for classical hardware while classical systems run the rest of the workflow. Drug discovery, materials modelling and financial optimisation all follow this pattern. Quantum Motion is building toward these applications through projects including QUICHE, a partnership with Riverlane and FACCTs developing end-to-end quantum chemistry workflows.
Hybrid computing only works well when the quantum and classical hardware are co-located, so that data moves between them without the latency of distance. It also needs to plug into the tools HPC teams already use, frameworks like Qiskit and schedulers like Slurm, rather than asking them to build a new operational stack around one machine. A quantum computer that fits in a standard rack and draws less than 200 kilowatts can sit on the same floor as existing infrastructure and use the same workflows. One that requires its own facility and hundreds of megawatts cannot do that. It becomes a competing demand on an already constrained grid rather than a complement to the systems beside it.
Scaling without sprawl
As silicon spin qubit systems scale to larger qubit counts, the design intent is for the chips to get more complex without the surrounding machine getting much bigger. Reaching a utility-scale, fault-tolerant processor with broad applications across drug discovery, materials science, finance and logistics should be possible within a footprint comparable to the system already deployed at the NQCC, and within the power demands of a standard data centre.
AI's energy demand caught most policymakers off guard. Quantum computing doesn't have to repeat that. The question of how much power a useful quantum computer will draw, and where that power will come from, is worth asking properly now, while there's still time for the answer to shape how the industry builds.
Key facts:
- Quantum Motion's target utility-scale system draws less than 200 kilowatts. Independent analysis estimates superconducting, ion-trap, and photonic systems at equivalent computational scale would require 160, 140, and 100 megawatts respectively.
- Quantum Motion's deployed system at the UK's National Quantum Computing Centre fits within three standard 19-inch server racks.