The study evaluates potentially significant energy and material requirements for future systems, without establishing that cooling negates computational savings. A study published in June 2026 by researchers affiliated with an American national laboratory quantifies this for the first time at large scale. The promise of clean, universal computational power meets a physical constraint that the industry had carefully avoided quantifying until now.
The Essentials
- The study concludes that energy and material costs would present significant challenges for planning, not that they demonstrably erase all energy gains.
- Cryogenic infrastructures are concentrated in a few regions of the world, but remote access via the cloud is gradually expanding to researchers, students, developers, and smaller organizations.
- Pasqal and the Open Compute Project are exploring architectures without extreme cooling, but none has yet demonstrated viability beyond the laboratory.
- The fork between concentration and gradual accessibility will play out over the coming years.
Cold as the Price of Entry
Cryogenic consumption presents a structural asymmetry: a base portion remains fixed, but active thermal loads vary with operation and control signals. Cooling maintains a baseline load, but it does not consume strictly as much at 10% capacity as at full capacity. This incompressible nature of the base cost changes the economic logic of deployment: amortization does not improve with intensive use in the same way as for a conventional supercomputer, where consumption is more proportional to actual activity.
A quantum superconductor operates at temperatures that the observable universe almost never reaches naturally. At 0.015 kelvin, roughly 273 degrees below zero, superconducting qubits maintain their quantum coherence long enough to perform calculations. Reaching this temperature in a laboratory setting requires dilution refrigeration systems that few institutions worldwide can afford, install, and maintain.
The McCollum et al. article presented in June 2026 a first forward-looking analysis of large-scale energy and material requirements. The study includes several resources mobilized by cooling, but not the full indirect impacts of the energy chain. The result is severe. The study addresses possible impacts of future large-scale deployments, not an established comparative assessment of already-deployed systems.
Energy assessments should integrate the entire infrastructure, but claims of savings must be examined case by case.
Quantum algorithms are theoretically promising for certain classes of problems, notably factorization, molecular simulation, and combinatorial optimization, but generalized practical superiority is not established for them. The cryogenic infrastructure that keeps them operational consumes continuously, whether the system is calculating or waiting.
The Gap Between Promise and Complete Accounting
Adam Becker, author of a documented critique of Silicon Valley techno-optimism, identifies a recurring mechanism: new technologies announce themselves in their best light, systemic costs remain out of frame until a third party measures them. Quantum computing offers a clear example. For ten years, actors in the field have communicated about computational power per qubit, error reduction, execution speed. Literature already contained energy models integrating cooling, but lacked integrated forward-looking assessments of resources at scale.
The CETEP study imposes a total-cost reading and requalifies the technology. The cryogenic infrastructure poses cost challenges for large-scale deployment. Proprietary operation of superconducting machines requires substantial infrastructure, but access can also be provided via the cloud, pending technological trajectory reducing this gap.
This reading competes with another, carried notably by economists like Philippe Aghion, who emphasize the nonlinear character of technological progress: breakthrough technologies always traverse a phase of infrastructural inefficiency before reaching industrial maturity. Early power plants were absurdly inefficient. Early internet servers consumed entire rooms for trivial capacity. The reasoning is not without force. But it assumes that the cryogenic physical constraint is itself solvable, which nothing guarantees within the decade.
The fundamental issue is not whether quantum computing is promising. It is. For well-documented use cases. The question is who can use it during the phase when infrastructure cost remains prohibitive.
Who Can Afford Cold
Geographic concentration of expertise tends to reinforce economic concentration: experts capable of maintaining these infrastructures are trained in the institutions that host them, making independent emergence elsewhere difficult. An actor wanting to build national capacity would need to simultaneously finance equipment, attract these experts, and train successors, unable to rely on a preexisting local industrial base. This path dependence amplifies the advantage of already-established actors beyond their mere capitalistic superiority.
Quantum infrastructures are owned by public bodies, companies, national laboratories, and certain universities, with highly unequal capacities. Access via cloud is expanding. IBM is deploying its quantum processors in specialized data centers accessible via the cloud. Google has announced massive investments in its quantum division. In China, several national laboratories operate comparable systems with direct state support.
In Europe, the European Union’s Quantum Flagship program mobilizes one billion euros over ten years to structure a sector.
This geography is not trivial. It reproduces the model of civil nuclear power or space: a technology whose mastery confers strategic advantage and whose infrastructure possession remains concentrated. End users access it via cloud, under pricing and regulatory conditions set by actors they do not control.
This model has stable precedents. Internet infrastructure is concentrated in the hands of a few actors whose undersea cables and data centers constitute critical nodes. The questions of digital sovereignty this generates remain unresolved. Quantum computing could repeat this structure while making it even more unequal, because the physical barrier of absolute zero is harder to circumvent than the cost of a fiber optic cable.
Current concentration is not solely economic. It is also geographic: expertise in superconducting quantum systems remains concentrated in several laboratories located in North America, Western Europe, China, and Japan. Many territories do not necessarily possess local infrastructure, but they can participate via remote access to quantum services.
Pasqal and the Path of Neutral Atoms
Facing the dominant superconductor model, alternative approaches exist. Pasqal, a French company spun out from the physics laboratory of the École Normale Supérieure, works on neutral atom qubits. These atoms, held in place by laser optical tweezers, do not need the extreme cryogenic temperatures of superconductors. They operate at still very low temperatures, but in a range achievable with much lighter and less energy-intensive equipment.
The architecture differs in its constraints: neutral atoms offer better qubit connectivity at short distance, but scaling up to thousands or millions of qubits poses engineering challenges that superconductors have already partly solved. Pasqal has deployed neutral atom systems in HPC and industrial environments, notably at Jülich, the CEA, and Aramco, and delivered a 140-qubit computer to CINECA. Their viability at very large scale remains to be demonstrated.
The Open Compute Project, a tech consortium founded by Meta to standardize data center infrastructures, became interested in the quantum question from the angle of modularity. If open standards allowed assembling quantum components as servers are assembled today, concentration of supply could decrease. But these standards do not yet exist, and their adoption would require dominant actors to accept renouncing part of their proprietary advantage.
These two paths, neutral qubits and modular standards, sketch a possible alternative to cryogenic monopoly. They are not yet answers. The same tension between technological promise and viability at scale is found in other sectors where laboratory demonstrators are slow to become industrial infrastructures.
The 2026-2032 Fork and Its Effects
The decade beginning is decisive for two distinct reasons.
The first is technical. IBM projects 200 qubits for Starling in 2029 and up to 2,000 qubits for Blue Jay from 2033 onward. D-Wave projects a system of 10 logical qubits in 2030, after a milestone of 181 physical qubits in 2028. Google ultimately aims for an error-corrected computer of roughly 1 million physical qubits, with no official date of 2030. At large scale, the study describes uncertain tradeoffs between efficiency, demand, and resource constraints, without establishing a determined economic or energy threshold.
The second reason is structural. If alternative approaches, notably neutral atom qubits, succeed in scaling up, they could reduce access barriers for certain public actors and universities. The costs and future trajectories of quantum computing are uncertain, and current projections from private actors integrate optimistic assumptions about resolving engineering challenges.
A concentration scenario would see quantum computing follow a model similar to critical technologies: possession concentrated in a few countries, cloud access for others, without immediate accessibility. A bifurcation scenario via neutral qubits could open a more distributed path, but would require coordinated public investments in open demonstrators and adoption of interoperability standards reducing dependence on proprietary suppliers.
The control of critical resources in the technology chain runs through all major industrial sectors. Quantum computing adds a physical constraint: the necessity of reaching absolute zero stems from the laws of thermodynamics, not a modifiable commercial rule.
Signals to Watch Until 2030
Two indicators will allow distinguishing trajectories before the fork is sealed.
The first is the number of operational quantum systems functioning without cooling below one kelvin, recorded in the technical roadmaps published by actors and national laboratories. If this number remains low after 2027, the dominant cryogenic model will have confirmed its predominance.
The second would be a ratio between cryogenic energy cost and computational savings generated, measured on real deployments and not laboratory benchmarks, but this assessment remains rarely publicly documented. Independent studies could replicate it on systems deployed in real conditions by cloud actors. If these measurements remain absent from public debate or confined to academic publications without industrial response, it is a signal that dominant actors prefer opacity to comparison.
These two signals do not say where the technology is going. They say where it actually is, and who in this landscape has interest in that being clearly known.
The European quantum program, if it wanted to weigh on this fork, should integrate net energy efficiency criteria into its calls for proposals, finance demonstrators of qubits without extreme cooling in open consortiums, and require total-cost performance reports for systems it finances. None of these conditions is imposed today. The window for setting them remains open, but it is narrowing as proprietary standards become established.
Sources
- CETEP/National Laboratory quantum infrastructure study, May 2026, TechXplore
- Adam Becker, critique of Silicon Valley techno-optimism, adam-becker.com
- UBS, report on the economics of quantum computing, January 2026 (no verified link)
- IBM Quantum roadmap 2025-2030, IBM Research
- Quantum Flagship program, European Commission, ec.europa.eu
- Pasqal, neutral atom qubits architecture, pasqal.com
- Open Compute Project, opencompute.org
