Quantum Computers Are “Always Five Years Away.” The Targets Are Not.

Introduction

On July 20, Christophe Jurczak, co-founder of Pasqal and managing partner of the quantum-focused venture firm Quantonation, published a paper in MDPI’s Philosophies journal arguing that quantum computing is a “Perpetual Five-Year Technology”: a class of physics-based technologies whose horizons reset each time a milestone is crossed. Eleven days later, New Scientist gave it the headline treatment: “Why full-fledged quantum computers might always be five years away.”

The paper builds on an earlier Quantonation whitepaper released in December 2025 that introduced the PFYT concept in an investor-facing format. The academic version is more rigorous, drawing on the French philosopher of technology Gilbert Simondon’s theory of “technical individuation,” which holds that a technology becomes a coherent object through mutual conditioning between its components and its surrounding ecosystem.

Jurczak defines four criteria for a PFYT: milieu-dependence (performance bottlenecked by multiple external subsystems), concretization dynamics (components must become mutually conditioning), recursive constraint discovery (each solved bottleneck reveals deeper ones), and high validation latency (iteration cycles are intrinsically slow). He argues that quantum computing meets all four, and that the perpetual five-year horizon is therefore not a forecasting failure but the “temporal signature” of how these technologies develop.

The paper’s empirical discussion centers on two datasets. Hardware-performance curves draw on the Riverlane Quantum Error Correction Reports (2024 and 2025 editions), showing error rates converging across six quantum-computing approaches (trapped ions, superconducting circuits, neutral atoms, silicon spins, photonic, and bosonic codes), with later-entering platforms appearing to improve more rapidly. A separate bibliometric dataset, built from OpenAlex records and compiled by Brooke Abeles, tracks the distribution of high-impact research across platforms. This reveals what Jurczak calls the “trapped-ion paradox”: ions achieve the best gate fidelities of any platform, yet their share of high-impact research has declined from approximately 15–19% in 2011–2012 to an average below 6% after 2017. His interpretation is that mature platforms with less room for conceptual innovation lose “epistemic momentum” to platforms in active development, currently neutral atoms.

Jurczak explicitly distinguishes PFYTs from pathological science: cold fusion, he notes, exhibited “perpetual deferral without genuine concretization” because no associated milieu accumulated. He also disclaims predictive power, framing the PFYT concept as diagnostic rather than forecasting.

The conflict of interest is disclosed in the paper: Jurczak is co-founder of both Quantonation and Pasqal, a neutral-atom quantum computing company whose systems are extensively referenced. Section 4, a philosophically framed case study of neutral-atom processors as exemplary technical individuals, centers on the technology modality with which he is professionally affiliated. The paper acknowledges this and states the case “reflects analytical clarity, not advocacy,” but the overlap between author and subject is worth noting.

Padavic-Callaghan’s New Scientist column presents the PFYT thesis largely unchallenged. Padavic-Callaghan reports being “increasingly won over” to Jurczak’s view without testing it against concrete engineering milestones of the past three years or pressing on the conflict of interest beyond a passing mention of the affiliations.

PFYT Names a Real Pattern, but Overextends It

Jurczak’s paper is the most intellectually interesting piece of quantum meta-analysis I have read this year. The PFYT concept names a pattern that anyone who has followed quantum computing for more than a decade recognizes: the five-year horizon does reset. Improvements in coherence shift attention toward crosstalk at scale. Lower logical error rates make decoder throughput and classical feedback the bottleneck. Reductions in one resource layer move the dominant cost to magic-state factories or cryogenic control. Progress does not simply remove constraints; it changes which ones matter most. Jurczak is right that this is not a forecasting failure. It is the structure of the problem.

The Simondon framework adds genuine value. Technical individuation (the idea that a technology becomes a coherent object through co-evolution with its surrounding ecosystem: cryogenics, control electronics, fabrication, workforce, theory) maps well onto what I observe through my CRQC Quantum Capability Framework. No single capability can be optimized in isolation because each conditions the others. The framework tracks ten capabilities across four tiers precisely because the mutual conditioning Simondon describes is an engineering reality, not a philosophical abstraction.

Jurczak’s strongest point is that the technical object remains unsettled: architecture, error correction, control, fabrication, decoding, and classical infrastructure are still co-evolving. But that does not make every arrival question ill-posed. A capability can have a stable acceptance test before the machine that will satisfy it has reached a stable technical form.

He writes: “to ask when a technology will ‘arrive’ presupposes that we already know what it is; if the identity of a quantum computer is still being settled as the machine develops, that presupposition fails.”

For the most general version of the question — what will quantum computing ultimately become? — that’s fair. Nobody knows whether the long-term dominant form will be superconducting, neutral-atom, photonic, topological, or some hybrid. Nobody knows which business models will work, or whether quantum computing’s greatest value will be in computation, sensing, or generating training data for AI systems. The heterogeneous future Jurczak sketches is plausible.

But quantum computing is not one question. It is several. And for some of them, the acceptance test is already concrete even while the architecture that will pass it continues to evolve.

A CRQC Has a Definable Acceptance Test

A cryptanalytically relevant quantum computer is operationally definable. Specify the target cryptosystem, the production key size, the required success probability, the acceptable runtime, and the question becomes one of resource estimation. Gidney’s 2025 preprint estimates that RSA-2048 could be factored using 1,399 logical qubits and roughly 900,000 physical qubits in under a week, under explicit surface-code assumptions. That is a major reduction from the frequently cited 2019 estimate of 20 million physical qubits, although the earlier design targeted a shorter runtime. Chevignard et al. estimated 1,193 logical qubits for the 256-bit prime-field elliptic curve at EUROCRYPT 2026, with 22 independent runs and a substantial Toffoli gate budget. My CRQC Quantum Capability Framework decomposes the engineering path into ten capabilities across four tiers, each with specific thresholds. The CRQC Readiness Benchmark lets anyone plug in their own assumptions and see how the estimates shift.

The functional target of this machine is settled enough for engineering, procurement, and policy decisions, even though the eventual hardware pathway (which modality, which error correction code, which control architecture) remains open. We can state the task and estimate resources under explicit assumptions. We cannot yet say when an integrated hardware stack will satisfy those assumptions at scale. The uncertainty is in the timeline, not the target. Simondon’s individuation framework may well describe the ongoing co-evolution of quantum hardware subsystems, but it does not make the CRQC’s acceptance criterion any less concrete.

Useful Fault-Tolerant Machines Are Also Definable

The CRQC is one target. Industrial utility is another, and it is also more defined than Jurczak’s framework suggests.

My Quantum Utility Map series spent seven articles surveying a broad body of published fault-tolerant resource estimates and mapping them to the real-world problems they address. The Quantum Utility Ladder that emerged from that analysis maps logical qubit counts to specific industrial capabilities: 25 to 100 logical qubits for small-molecule energetics and NMR prediction, 100 to 300 for OLED materials and battery cathode candidates, 300 to 1,000 for cytochrome P450 metabolism and CO₂ catalyst screening, 1,000 to 5,000 for full FeMoco and bulk electrolyte simulation, and beyond 5,000 for inertial confinement fusion modeling and lattice gauge theory.

These are not vague aspirations. Individual studies specify circuit depths, qubit counts, gate budgets, and error rate requirements. A pharmaceutical company can, in principle, specify the molecular system, the required observable, the precision, and the economic comparator. That defines the capability threshold far more concretely than the phrase “useful quantum computer,” even though it does not produce a reliable calendar date. The identity of “a quantum computer useful for screening transition-metal catalysts” is not in a state of Simondonian becoming. It is an engineering specification with quantified requirements.

The applications where quantum advantage is most plausible (pharmaceutical R&D, chemical catalyst design, battery materials, advanced materials science) have the strongest structural case because generic exact simulation of quantum-mechanical systems can scale exponentially on classical hardware while appropriate quantum algorithms offer more favorable asymptotic behavior. Current end-to-end resource estimates tend to look less favorable for generic optimization, finance, and machine-learning workloads, particularly once data loading, oracle construction, fault-tolerance overhead, and continuously improving classical baselines are included. As I argued in Quantum Computing by 2033 and in Quantum Chemistry’s Honest Ledger, these are grounds for skepticism about the latter category, not a universal impossibility result, but the asymmetry is pronounced.

Defined Capabilities Can Emerge Before Architectures Settle

The paper’s most interesting evidence also places a limit on its own thesis. Section 3.2 discusses two recent quantum simulation validations: a 256-atom neutral-atom simulator reproducing the frustrated quantum magnet TmMgGaO₄ in quantitative agreement with laboratory measurements, and a superconducting digital quantum processor computing the dynamical structure factor of KCuF₃ benchmarked against neutron-scattering data.

Jurczak cites these as evidence of “PFYT dynamics” because they appeared nearly simultaneously on different platforms. His conclusion: “the ecosystem has matured to the point where simulation capabilities emerge cross-platform, but the plurality of computational paradigms persists rather than converging.”

I read the same data through a practical rather than philosophical lens. Jurczak sees cross-platform co-existence as evidence of unsettled identity. A domain scientist sees two different instruments delivering validated, quantitative results for real physical systems. The Pasqal preprint reported quantitative agreement for magnetization and phase behavior in TmMgGaO₄, alongside comparisons of correlations and structure factors. The superconducting preprint computed the dynamical structure factor of a one-dimensional spin chain and benchmarked it against matrix-product-state calculations and neutron-scattering measurements. Neither result, standing alone, demonstrates fault-tolerant computation or commercial quantum advantage. But they are credible, quantitatively testable demonstrations of defined scientific capability.

A quantum platform does not need to have reached its final Simondonian form before a user can define a job, run it, and evaluate the result. The identity question Jurczak raises — analog simulator? digital gate processor? photonic medium? — is real at the level of the ultimate architecture. It is irrelevant to the researcher who just got a validated simulation. Technical identity can remain unsettled while capability thresholds become concrete.

The Architecture Is Unsettled. Some Thresholds Are Not.

Jurczak is right that the ultimate architecture and economic role of quantum computing remain open. He is wrong only if that openness is allowed to swallow every narrower question.

We can define an operational threshold for attacking RSA-2048 without knowing which modality will reach it. We can define the observable, precision, and runtime required for a useful simulation without knowing the ultimate form of the processor. And we can recognize a validated scientific result without pretending that it establishes general-purpose utility. The architecture may continue to individuate. The acceptance tests can nevertheless become concrete. PFYT explains why the route keeps changing. It does not make every destination unknowable.

In 1995, nobody could predict that the internet would produce Instagram, Uber, or Bitcoin. But anyone could see that a global network would be useful for commerce, communication, and research. Unknown downstream applications do not make the underlying capability undefined. Quantum computing is in the same position.

The New Scientist Problem

Jurczak’s paper, with all its blind spots, is a serious piece of work. The New Scientist headline is something else.

“Why full-fledged quantum computers might always be five years away” converts a diagnostic account of development dynamics into a prediction of indefinite non-arrival. That is not what Jurczak claims. PFYT is intended to explain why forecasts reset as the technical object develops; it is not evidence that quantum computers can never pass defined capability thresholds. The paper explicitly draws a line between PFYTs and pathological science. Its own data shows real, measurable progress across every platform. None of that made it into the headline, and readers who see only the headline will take away exactly the wrong conclusion.

The coverage would have been stronger had it tested the philosophical thesis against recent platform-specific milestones: Google’s below-threshold quantum error correction result, the 448-atom experiments implementing key elements of a universal fault-tolerant neutral-atom architecture, the quantum simulation validations Jurczak himself cites, real-time decoder demonstrations, and cross-platform error-rate convergence. The horizon may reset, but the ground covered does not disappear. Padavic-Callaghan never engages with that distinction.

Jurczak’s ENIAC comparison requires more precision than either the paper or the coverage gives it. Jurczak is right that classical computing possessed a more stable computational abstraction before reliable hardware arrived: Boolean logic and Turing’s mathematical framework were settled before ENIAC was built. That is a fair point. But it does not establish that an unsettled implementation pathway makes every capability threshold undefined. Shor’s algorithm at production key sizes is as well-defined an acceptance test as Boolean arithmetic was in 1945, even though the machine that will pass it has not yet been built. The quantum computing abstraction is unsettled for the most general case; for production-scale cryptanalysis and some simulation workloads, the abstraction is already here.

Jurczak discloses his roles at Quantonation and Pasqal, and Section 4 centers on the neutral-atom modality with which he is professionally affiliated. That context does not invalidate the analysis. It does make independent scrutiny of platform selection and evidentiary framing particularly important, and the New Scientist column did not provide it. The paper itself handles the conflict of interest with considerably more rigor than the coverage it received.

What the PFYT Framework Gets Right, and What to Do With It

For all its blind spots, the PFYT framework has genuine utility for my readers, whether you are investing in quantum companies, setting national quantum strategy, building PQC migration programs, or evaluating vendor roadmaps.

The recursive constraint discovery pattern is real and worth internalizing. When a quantum hardware vendor announces a milestone, the right question is not “so are we there yet?” but “what constraint did this reveal?” Every qubit-count record, every fidelity improvement, every error correction demonstration opens a new bottleneck. Understanding this prevents both irrational exuberance (one result does not mean the machine is around the corner) and irrational dismissal (progress is real even when the horizon resets).

Jurczak’s observation that later-entering platforms appear to improve more rapidly on error-rate curves, a pattern he attributes partly to inherited ecosystem knowledge, is useful for investors and policymakers. Evaluating a platform solely by its current performance is misleading. A neutral-atom system that is two years behind superconducting circuits on gate fidelity may be on a steeper trajectory because it inherits solutions to problems the superconducting community spent a decade solving.

The PFYT framework is also a useful corrective to vendor roadmaps. If each milestone genuinely reveals the next constraint, then linear-extrapolation roadmaps (“we’ll have 100 logical qubits by 2028, 1,000 by 2030, 10,000 by 2032”) deserve skepticism not because the company is lying but because the milestones they haven’t hit yet will reveal constraints they haven’t anticipated. This is not an argument for inaction. It is an argument for building flexibility into plans: crypto-agility in PQC migration, modular architectures in quantum hardware procurement, scenario-based rather than point-estimate planning.

But none of this changes the operational picture. The PFYT debate is intellectually interesting. It changes nothing about the decisions facing my readership.

For CISOs and CTOs, the policy clock is already running. FINMA recommends a board-backed PQC roadmap by mid-2027. The UK NCSC targets estate discovery and planning by 2028, priority migration by 2031, and completion by 2035. The EU coordinated roadmap calls for transition activity to begin by the end of 2026, with high-risk systems migrating by 2030. CNSA 2.0 sets more specific transition requirements for U.S. national-security systems. These are not one homogeneous set of hard deadlines, but together they make waiting for architectural closure an indefensible strategy.

Investors should find the PFYT framework useful because it shifts diligence away from headline qubit counts and calendar promises. The relevant questions are which constraint the latest milestone has exposed, whether the proposed architecture can absorb that constraint, and whether the roadmap integrates error correction, control, decoding, interconnects, and workload-level economics rather than extrapolating one metric.

For policymakers and national strategists, the argument that quantum computing is “always five years away” is precisely the kind of framing that leads governments to defer investment and then find themselves dependent on foreign quantum infrastructure when the machines arrive. My analysis in Quantum Sovereignty documents what happens when nations treat quantum computing as a future problem rather than a present one: they end up on the wrong side of the utility trap, dependent on whoever controls the quantum supply chain for their pharmaceutical, chemical, and materials research. Whether the general-purpose horizon resets is irrelevant to that strategic calculus.

The quantum computer does not need to acquire a final philosophical identity before it becomes operationally consequential. The architecture can remain open while cryptographic risk, scientific capability, and national dependency become concrete. The moving horizon is a reason to plan flexibly. It is not a reason to wait.