Balancing the “if” with the “when”
A wave of optimism is building around the prospects for achieving a usable, fault-tolerant quantum computer, propelled by impressive development progress over the last 12 months. However, in reality, this goal remains some distance away, and major technological hurdles still need to be overcome. Is the optimism wholly justified?
In this update to Arthur D. Little’s (ADL’s) Blue Shift study from three years ago, we look at what has been achieved, what is now being promised, and how executives can stay critical as they assess future progress claims.
A WAVE OF OPTIMISM
In September 2022, amid growing optimism around the progress of quantum computing (QC) technology development, we released the Blue Shift report “Unleashing the Business Potential of Quantum Computing.” The report focused on computing applications, excluding other quantum technology areas such as communications, sensing, and cryptography. Its goal was to equip business executives with a clear view of the current landscape, insight into the growing ecosystem of players, and a perspective on commercialization timescales and probabilities. More than three years later, it’s worth reflecting on what has — and hasn’t — been achieved.
Despite the inherent complexity of QC technology, it might seem relatively straightforward to form a reasonable view of its current business prospects. There’s certainly no shortage of technology analysts and commentators, and most R&D is still in the public domain. Yet, in reality, forming a clear, unbiased picture is difficult. Looking at the scale of public investment — with governments in the UK, China, Europe, and the US committing many billions of dollars — along with further billions from private investors and renewed calls for action from think tanks such as the Tony Blair Institute in the UK — one might conclude that the outlook is overwhelmingly positive.
Nearly every month, development players announce new milestones toward scale-up and commercialization — many of which are genuinely impressive. In January 2025, MIT Technology Review stated that “useful quantum computing is inevitable — and increasingly imminent.” Later that year, Google CEO Sundar Pichai remarked, “I would say quantum is where AI was five years ago. So, I think, in five years from now we’ll be going through a very exciting phase in quantum.” Today, most influential commentators suggest that — despite remaining technical challenges — the commercialization of QC is a matter of “when,” not “if.”
But does this message tell the whole story? In this Viewpoint, we argue that executives aiming to position their businesses for a QC future must continue to take a critical and measured approach. We briefly review key developments in QC over the past few years, explore future prospects, and highlight several reasons why caution remains warranted.
REASONS TO BE CHEERFUL
The past three years have seen significant progress in the long journey toward a practical fault-tolerant quantum computing (FTQC) device. An FTQC would be applicable across industries and could provide a meaningful advantage over conventional high-performance computers (HPCs) for particular classes of intractable problems — that is, problems for which the solution time scales exponentially with problem size. Figure 1 summarizes some of the most noteworthy milestones since 2022.
Three years ago, the QC race primarily focused on scaling the number of physical qubits and advancing error-correction techniques — combining hundreds or thousands of qubits to create error-corrected “logical qubits.” At the time, there were some hopes that noisy intermediate-scale quantum (NISQ) devices could be scaled further to produce a usable quantum computer. While many NISQ-era quantum algorithms are close to bringing some quantum advantage, most largely focus on fundamental physics simulations and remain far removed from the practical needs of industry users.
As a result, most current efforts lean toward the development of FTQC. Generally, to be usable in a broad range of applications, an FTQC would require at least 100 logical qubits, with many of the more valuable applications demanding thousands. This, in turn, implies that the number of physical qubits needed would be in the thousands to millions, depending on the underlying technology. To date, however, experimental systems have only reached the scale of hundreds of physical qubits.
Four underlying quantum technologies are currently being pursued for gate-based quantum computers. Superconducting is arguably the most mature, followed by trapped ion, neutral atom, and photonics. Other technologies, such as electron spin, are also being pursued but are less mature (not shown in Figure 1).
Progress has accelerated considerably over the last 12 months, with notable achievements including Quantinuum/Microsoft’s significant improvements in repeatable error correction, logical qubit stability, and gate fidelity, as well as Google’s Willow 105-qubit chip, which demonstrated in 2024 that increasing the size of error-corrected qubits can reduce the overall error rate — key for the viability of scaling up. In December 2025, Google also implemented magic state cultivation, based on an idea published in 2024, which enables the synthesis of a logical T gate, a cornerstone feature for achieving exponential speedups in quantum algorithms. However, much remains to be done in implementing logical quantum gates and real-time fault tolerance at a scale requiring tens of thousands of high-quality physical qubits. IBM’s Nighthawk processor, presented in November 2025, demonstrated a significant advance in qubit connectivity, enabling lower-overhead error correction and showing that existing, affordable technology could scale with quantum processor development.
Neutral atom technology posted a gain in progress toward scale-up in mid-2023 through Atom Technology’s array of more than 1,000 qubits (the largest achieved so far), although the array was not actively controlled for processing. Between December 2023 and November 2025, QuEra posted significant breakthroughs in error correction, scalable codes, and other key enablers for building a future FTQC. On the photonics side — a less mature but fundamentally different development path focused on fault tolerance from the outset — PsiQuantum, in collaboration with GlobalFoundries, has demonstrated the capability to mass-produce quantum chips based on optical quantum computing. Quandela also demonstrated its ability to efficiently create deterministic entangled photons at scale.
Quantum annealers work on a different principle from universal gate-based computing devices and have more limited applications as special-purpose optimizers and samplers in areas such as logistics, financial simulations, and operations research.
Although their application is more limited, annealing-based quantum computers are, in many respects, the most mature QC technology when it comes to real-world use cases. In 2025, D-Wave announced significant advances in qubit connectivity, coherence, and noise reduction with the release of its Advantage2 system. The company claimed some quantum advantage in March 2025 for a specific hard spin-simulation problem, although this has been debated.
There is also progress in the viability of hybrid quantum/conventional approaches. For example, in June 2025, IBM and RIKEN in Japan demonstrated such a hybrid solution running on IBM’s Heron quantum computer and RIKEN’s Fujitsu Fugaku supercomputer to perform electronic-structure simulations of the ground-state properties of [2Fe-2S] and [4Fe-4S] clusters, using quantum circuits with up to 77 physical qubits and a record 10,570 quantum gates. Likewise, in October 2025, Nvidia announced NVQLink — a high-throughput, low-latency interconnect for quantum processing units, enabling FTQC vendors to implement real-time error correction.
Finally, there is new potential for the integration of AI with QC. In theory, AI has the potential to help tackle some of QC’s most difficult challenges. On the quantum engineering side, it could improve hardware calibration, error mitigation/control, and material/component discovery. On the software side, it could improve algorithm design, optimization, and data analysis. In the other direction, applications of quantum processing to improve AI are more speculative, but in principle, its parallel processing capabilities could accelerate AI training and inference, potentially helping reduce its very significant energy consumption. The main obstacle is the massive data-loading requirement, which is inefficient on quantum computers. This can be addressed more efficiently if the processed data is already quantum, such as data generated by quantum sensors. AI/QC integration remains at an early stage.
WHAT’S PROMISED
Looking forward, published roadmaps by key vendors promise a busy development schedule (see Figure 2).
On the superconducting development path, one of the most striking promises is IBM’s Starling FTQC, targeted for 2029. IBM claims it “really [has] a path to make this viable in this timescale.” Starling would have 10,000 physical qubits and 200 logical qubits, limiting its general applicability. IBM’s next device, Blue Jay, scheduled for 2033, would have 2,000 logical qubits, which is at the low end of what is needed to solve serious optimization and modeling problems. IBM forecasts that genuine quantum advantage could be achieved as early as 2026. Meanwhile, Google has avoided publishing specific device promises and instead describes a set of development milestones: M3 to M6. These milestones suggest a first fault-tolerant device around 2030, with the first broadly applicable FTQC arriving sometime between 2033 and 2035, roughly a decade from now.
Among trapped-ion approaches, IonQ has set the most ambitious targets, announcing plans for a 1,600 logical qubit device by 2028, followed by 8,000 and 80,000 logical qubit systems in 2029 and 2030, respectively.
For cold/neutral atom, development is at an earlier stage. Nevertheless, Pasqal is promising an FTQC with 200 logical qubits by 2029, and Infeqtion/Coldquanta are scheduling a device with “thousands of logical qubits” (100,000 physical qubits) by 2030.
Finally, for photonic qubits, PsiQuantum plans to deploy 100 logical qubits by 2028 into 2030, with a huge system consuming 100 MW and requiring cryogenics at the scale of the CERN Large Hadron Collider. Quandela plans for a similar outcome by 2029 but with a much smaller resource footprint.
Overall, the past 12 months have brought a remarkable wave of development across multiple vendors and technology platforms. The promise is that by around 2030, an FTQC with at least 100 logical qubits supporting around 1 million gate operations could emerge, enabling applications beyond highly specialized niches.
Toning down the hype
So far, the trajectory appears promising — but it is worth stepping back to reflect on the reliability of current predictions and whether the prevailing optimism warrants more nuance. Several important caveats need to be considered.
Timelines for a generally usable QC remain long
As we’ve seen, even the most optimistic forecasts suggest that the first prototype FTQCs are still five or more years away, with others, such as Google, projecting even longer time frames. Moreover, even once the first FTQCs emerge, it may take several additional years before fully engineered, commercial-grade devices suitable for general use become available and capable of solving everyday business problems. From the perspective of business end users outside the R&D function, technologies that may only become available seven to 10 years from now are still beyond normal planning cycles.
Significant scientific/technical risks & uncertainties remain
For nonexpert outsiders, assessing the specific technical risks involved with any of the current QC development paths is extremely difficult, if not impossible. What is notable in current roadmaps, however, is that perhaps the most fundamental technical challenge — scaling from hundreds to thousands and millions of qubits — has yet to be addressed. Progress to date, while impressive, has not yet proven beyond a reasonable doubt that this level of scalability is achievable. Reaching it depends on several unresolved uncertainties: the ability to manufacture various chip technologies at scale with high quality and low variability; the potential emergence of new sources of noise at larger scales; and the development of reliable quantum interconnects between quantum processing units, given the inherent limits on the number of physical qubits.
Practical FTQC applications may be more limited than many have assumed
Some experts have questioned whether the theoretical speedups promised by quantum algorithms can be realized in practice on future FTQCs, due to factors such as slow gate speeds and the very high overhead required for error correction. Lower polynomial speedups, such as quadratic ones, are generally considered insufficient to deliver practical quantum advantage over conventional devices, while exponential speedups may only be possible for certain types of problems. Relatively slow data loading and the current absence of so-called qRAM (quantum random access memory) constitute a barrier to applying QC to machine learning data analytics problems.
Taken together, this means that FTQCs are likely to be valuable mainly for very hard, highly structured problems with small outputs and enormous classical runtimes, such as simulations in chemistry, materials science, and high-energy physics. Their value for solving generic optimization problems, as well as for data analytics and operational applications requiring low latency, is highly dependent on the outcome of future fundamental research in quantum algorithm design. Governments and policymakers therefore must continue financing academic research in this and related domains, as well as funding schemes for more mature technology developments (e.g., through public procurements targeting industry vendors).
Past performance shows frequent milestone delays
It’s helpful to look at how the QC industry has fared in delivering on previous roadmap milestones. Even without conducting a comprehensive analysis, it’s fair to say that performance has been mixed. For example, at the time of ADL’s last study in 2022, IBM had roadmapped a “quantum-centric supercomputer” for about 2025, which has since been given a much longer time horizon. Similarly, Rigetti had targeted scaling to 1,000+ qubits by 2025, a milestone now postponed to 2027. Pasqal also envisaged 1,000+ qubits by the mid-2020s, but this has since been replaced by a more conservative plan. These examples reflect the continuing challenges associated with scaling, error correction, and control complexity.
The QC ecosystem has vested interests in sustaining optimistic messaging
The ecosystem of developers, suppliers, researchers, investors, and government bodies needs to build confidence and public support to maintain and further grow its activities. While the vast majority of players presumably act in good faith, the pressure to generate positive publicity is often overriding. With both investors and developers holding substantial stakes — and large amounts of capital already committed — there is limited incentive to question overly optimistic narratives. Unfortunately, it is not always possible to solve a problem simply by throwing money at it.
QC information in the public domain is often noisy & misleading
The complexity of QC — both in its engineering and in its potential application modes — makes it easy for nonexperts to misinterpret the true significance of reported milestones and optimistic projections about its potential benefits. In his comprehensive and continuously updated book Understanding Quantum Technologies, quantum engineer Oliver Ezratty highlights many examples of misunderstandings and misconceptions regarding progress and challenges in QC. At a basic level, predictions of future QC market sizes are sometimes conflated with estimates of end-user business value creation, resulting in a greatly inflated impression of the potential market. A common, but misplaced, belief is that quantum computers will replace HPC systems across many applications. In reality, QC is far more likely to complement HPC, particularly for targeted problem classes such as chemistry simulations. Finally, despite some claims to the contrary, it is unlikely that QC will ever enable large-scale big data applications.
Classical computing is also making progress
The advent of quantum-inspired techniques running on classical computing resources, including increasingly powerful dedicated GPUs such as those from Nvidia, as well as advances in classical deep learning methods, is continuously moving the goalposts for achieving QC advantage. This further reinforces the need for end users to monitor and assess a broad range of technology trends and developments, not only in QC, but also in classical and emerging paradigms such as photonic, reversible, and probabilistic computing.
CONCLUSION
Stay tuned & evaluate progress critically
In our 2022 Blue Shift report, we outlined four key actions for executives to avoid being left behind by potential QC breakthroughs: (1) explore applicability, (2) monitor technology developments, (3) engage with the ecosystem, and (4) build knowledge and capability. These steps remain just as relevant today.
Ezratty proposes a valuable framework for critically evaluating emerging QC demonstrations, case studies, and use cases. Building on his work, we propose the following five evaluation criteria:
We do not seek to diminish the tremendous achievements being made in QC, which continues to hold immense transformative potential. The recent acceleration in development has undoubtedly brought us closer to the long-sought goal of a broadly applicable, universal gate-based quantum computer. However, it is easy to be swept up in the surrounding hype — perhaps even more so in the wake of AI’s meteoric rise. It is worth remembering that modern AI development began more than 60 years ago and went through multiple periods of disillusionment before achieving today’s breakthroughs. QC is still far less mature than AI. In QC, patience remains a virtue, particularly as the “if” has not yet fully become a “when.”
With kind acknowledgment of contribution from Olivier Ezratty, quantum engineer
cloud technology
Este fondo, con un tamaño objetivo de 250 millones de euros, pretende...
