The Quantum Computer That Is Always Arriving
IBM has built some remarkable machines. Its greatest achievement may yet be persuading the world that tomorrow is a product category.
Keywords: quantum computing, IBM, Qiskit, quantum simulation, classical computing, quantum advantage, fault tolerance, logical qubits, technological hype, Nighthawk, Heron, Starling, quantum-classical computing
There are industries in which one sells a product.
There are industries in which one sells the promise of a product.
And then there is quantum computing, where the product, the promise of the product, the simulator of the product, the development environment for the product, the roadmap towards the product, the conference announcing the next version of the roadmap, and the consultancy explaining why your board urgently requires a strategy for the product have all acquired separate invoices.
It is, in its way, exquisite.
The Victorians generally built the railway before charging for the journey. They were primitive people, burdened by the vulgar assumption that transportation required movement.
Modern technology has freed itself from such intellectual limitations.
Today one may establish the timetable, sell first-class tickets, launch a developer ecosystem, announce strategic partnerships, publish an educational curriculum and release a simulator enabling customers to experience something resembling the journey before anybody has built the train capable of reaching the advertised destination.
Progress is marvellous once one abolishes arrival.
Quantum computing has become especially good at this because the subject is sufficiently difficult that almost any objection can be made to sound provincial.
Ask whether a quantum computer performs a commercially valuable calculation better than an available classical machine and somebody will explain Hilbert spaces to you.
Ask again and one receives entanglement.
Persist, and error correction appears.
Continue beyond that and one is accused of failing to understand the transformative character of the technology.
At no point is it considered polite to return to the original question.
IBM is a particularly interesting case because the cheap criticism of it is false.
IBM does have quantum computers.
Real ones.
Its current hardware portfolio includes physical quantum processors such as Heron and Nighthawk. IBM describes Heron systems with 133 or 156 programmable qubits and Nighthawk processors with 120 programmable qubits. Quantum System Two combines quantum processors with cryogenic infrastructure, modular qubit-control electronics and classical runtime servers. These are physical machines manipulating genuine quantum systems. They are not merely PowerPoint presentations with chandeliers attached.
That distinction matters because satire becomes tedious when reality has already supplied better material.
The serious criticism is not that IBM possesses no quantum hardware.
It is that the expression quantum computer has become elastic enough to cover an extraordinary range of things, from noisy experimental processors that unquestionably exist today to large-scale fault-tolerant machines that IBM itself still describes as future systems.
The difference is not semantic decoration.
It is the entire argument.
IBM currently says that its first large-scale fault-tolerant quantum computer, Starling, is planned for 2029. The company envisages Starling operating with 200 logical qubits and executing 100 million quantum gates. Beyond that, its roadmap places Blue Jay in the 2033-and-beyond period, with ambitions extending to 2,000 qubits and one billion gates. IBM explicitly labels the roadmap as a statement of present intentions and goals subject to change or withdrawal.
In other words, the quantum computer that most members of the public think has already arrived is, according to IBM’s own roadmap, still scheduled.
There is something beautifully modern about selling access to the future while correctly footnoting that the future remains prospective.
IBM has physical quantum processors now.
It aims to have large-scale fault-tolerant quantum computing later.
Between those two statements lies an industry worth rather a lot of money.
One should admire the engineering.
One should admire the physics.
One should admire the enormous intellectual achievement required to manipulate fragile quantum states under extraordinarily difficult experimental conditions.
One is not therefore obliged to admire the grammar.
The central problem in quantum computing is rarely whether quantum mechanics is real.
It is.
Nor is the problem whether engineered quantum devices can manipulate qubits.
They can.
Nor is the problem whether these devices perform calculations.
They do.
The irritating question is far less metaphysical.
What useful computation can a quantum system perform sufficiently accurately, sufficiently reliably, sufficiently cheaply and sufficiently better than the best practical classical alternative that a rational user should prefer it?
There.
The vulgarity has entered the room.
Economics has always had this unfortunate effect on physics.
A physicist may demonstrate something astonishing.
An engineer may make it repeatable.
Then an economist asks what it costs and everybody suddenly remembers another appointment.
This distinction between possible, demonstrated, useful, advantageous and commercially superior matters because technological history is littered with magnificent machines that were successful at everything except becoming what their publicity predicted.
Quantum computing should be evaluated more strictly precisely because it is serious science.
It should not require rhetorical subsidies.
IBM itself has recognised the importance of distinguishing different stages of capability. Its public material has used concepts including quantum utility, quantum advantage and quantum-centric supercomputing rather than pretending that all quantum computation is equivalent. Its current Quantum Advantage Tracker explicitly exists because quantum and classical methods continue to compete, with apparently favourable results capable of changing as classical methods improve.
That is scientifically admirable.
It is commercially awkward.
A press release desires a finish line.
Science keeps moving it.
Classical computing is particularly inconsiderate in this regard.
Every time somebody discovers a calculation that looks hopelessly expensive classically, some ghastly person improves the classical algorithm.
Tensor-network methods improve.
Approximation methods improve.
GPU implementations improve.
Memory management improves.
Specialised numerical techniques improve.
Algorithms that were dismissed as impractical acquire optimisations.
The quantum machine may advance magnificently, only to discover that the classical machine has been training while everyone attended the launch party.
IBM’s Quantum Advantage Tracker is fascinating for precisely this reason. The company describes leading classical and quantum methods as being close enough that results can change as new techniques arrive. IBM specifically presents the tracker as a continuing process of comparison, correction and refinement rather than a ceremonial declaration that one side has permanently defeated the other.
The horse race is therefore not between “old computers” and “future computers.”
It is between two moving research frontiers.
This is where the public language of revolution becomes faintly comic.
If a quantum processor wins against one classical implementation on Tuesday, and a better classical technique recovers the ground on Thursday, civilisation has not moved backwards for forty-eight hours.
A benchmark is evidence.
It is not a coronation.
IBM’s July 2026 announcement on quantum advantage illustrates the seriousness of this difficulty. IBM described three lines of work intended to demonstrate quantum advantage while addressing the problem of validating answers once direct classical verification is no longer feasible. The company also stressed that advantage claims remain subject to comparison through its Quantum Advantage Tracker.
That qualification is not weakness.
It is good science.
It becomes inconvenient only after the public-relations department has already ordered the bunting.
Then there is simulation.
And here the comedy becomes almost indecent.
IBM’s own educational material teaches users to run quantum circuits either on real IBM quantum hardware or through the AerSimulator. IBM says explicitly that AerSimulator simulates a quantum circuit on a classical computer. Its documentation recommends simulators for development, testing, debugging and education before programs are sent to actual quantum hardware.
There is nothing wrong with this.
Simulation is fundamental to science.
Engineers simulate wings before building aircraft.
Physicists simulate stars without putting one in the laboratory.
Meteorologists simulate hurricanes without destroying Florida as part of quality assurance.
Nobody remotely competent objects to simulation.
What becomes absurd is the occasional rhetorical mist through which simulation, quantum computation and useful quantum hardware are allowed to merge into one glowing conceptual object.
A classical computer simulating a quantum circuit remains a classical computer.
It has not become quantum through sympathy.
It has not acquired superposition by installing a Python package.
It has not crossed an ontological boundary because the function name contains the word QuantumCircuit.
It is performing classical operations to calculate the behaviour prescribed by a quantum model.
A classical computer simulating rain does not become wet.
A computer simulating nuclear fusion does not require radiation shielding.
A machine modelling the Battle of Waterloo does not become Napoleon.
One would imagine these points too obvious to state.
Modern technological marketing has created a sizeable profession from things too obvious to state.
Qiskit’s documentation is admirably straightforward about the distinction. IBM provides local testing modes, fake backends based on QPU snapshots and Aer simulation. Fake backends can reproduce characteristics of particular IBM QPUs, including coupling maps, basis gates, qubit properties and noise information. Aer provides higher-performance classical simulation, including multiple simulation methods and custom noise models.
Again: excellent engineering.
Also: classical computing.
One begins to appreciate the peculiar position of the classical computer in the quantum revolution.
It is apparently obsolete, yet continually required to explain what its successor is doing.
The classical machine designs the quantum circuits.
The classical machine compiles them.
The classical machine simulates them.
The classical machine optimises them.
The classical machine helps control the hardware.
The classical machine processes measurement results.
The classical machine evaluates benchmarks.
The classical machine participates in hybrid workflows.
And after performing all this labour, it is invited to a conference at which somebody announces the end of classical computing.
Servants have always suffered from inadequate representation.
The irony becomes richer when one remembers why quantum computers are interesting in the first place.
The state space of a general quantum system grows rapidly with the number of qubits, making straightforward exact classical simulation extremely expensive. The very computational burden that motivates quantum machines also makes sufficiently large general-purpose quantum simulations difficult on classical hardware.
So we arrive at a wonderful circle.
We desire quantum computers partly because classical simulation of quantum systems becomes difficult.
We build classical simulators of quantum computers.
The simulators become expensive as the quantum systems grow.
We then celebrate the difficulty of the classical simulation as evidence that the quantum device has crossed an important computational boundary.
There may be impeccable complexity theory in the middle of this.
There is nevertheless something aesthetically satisfying about making a classical computer exhaust itself imitating the machine advertised as its replacement.
The CPU is invited to its own funeral and asked to operate the projector.
There is another distinction that publicity often treats rather carelessly: the difference between performing something classically difficult and solving something economically useful.
These are not the same achievement.
One can construct a mathematical problem of exquisite difficulty which nobody needs solved.
Difficulty does not create demand.
A lock may be extraordinarily difficult to pick.
That does not establish a market for the cupboard.
Quantum advantage is therefore important without automatically implying quantum usefulness in the ordinary commercial sense.
A quantum machine may demonstrate access to a computational regime beyond practical exact classical reproduction and thereby establish something profound about models of computation.
That is a major scientific achievement.
It does not follow that HSBC should migrate payroll.
One expects this distinction to be obvious.
One also expects annual reports to contain prose.
Hope is frequently disappointed.
IBM’s current hardware deserves better than hyperbole because the real engineering is impressive enough.
Nighthawk r2, announced on 31 August 2026, has 120 programmable qubits, 218 couplers and 120 reset elements. IBM reports more than 100,000 circuits per second, up to twenty-five times the circuit throughput of Heron, together with accurate results on circuits containing more than 7,500 gates. IBM presents the system as a substantial milestone in its 2026 roadmap and as useful for work related to error correction and eventual fault tolerance.
That is concrete progress.
Notice how satisfying concrete numbers are.
They have the wonderful property of not requiring the word revolutionary.
One may ask how many qubits.
One may ask how many gates.
One may ask what fidelities were obtained.
One may ask how the device behaves.
One may ask what error rates remain.
One may ask what workloads can be executed.
Numbers are horribly democratic.
They permit questions.
The trouble begins when a precise engineering accomplishment is inflated into an historical epoch.
A processor improvement becomes an era.
An era becomes transformation.
Transformation becomes inevitability.
Inevitability becomes procurement.
The adjective performs the work that the machine has not yet been asked to perform.
This is not peculiar to IBM.
It is a general pathology of technological industries.
Artificial intelligence has the same problem.
Blockchain had the same problem.
Nanotechnology had it.
The metaverse attempted it with considerable courage.
A field begins with genuine research.
Research generates difficult technical achievements.
Difficult achievements attract investment.
Investment requires a narrative.
The narrative must be larger than the engineering because nobody receives a billion-dollar valuation for “interesting specialist equipment with uncertain commercial scope.”
Thus progress is placed upon stilts.
IBM is actually unusually useful for separating the engineering from the mythology because its own roadmap remains explicit about the destination.
The company’s 2026 roadmap speaks of demonstrating initial examples of quantum advantage using quantum hardware together with high-performance computing. It envisages further scaling, quantum-classical workflow accelerators and, eventually, large-scale fault-tolerant systems.
Read carefully, the picture is not “quantum replaces classical.”
It is “quantum becomes integrated with classical.”
This is a much more plausible technological story.
It is also rather less cinematic.
IBM calls the approach quantum-centric supercomputing. Its hardware description for Quantum System Two explicitly includes classical runtime servers supporting hybrid workflows. The architecture envisages problems being divided between quantum and classical resources according to their respective strengths.
This may very well be where the practical future lies.
Quantum processors may become specialised accelerators within much larger computational environments.
They might handle particular tasks for which their physical properties provide genuine advantages, while CPUs, GPUs and high-performance computing systems handle everything else.
There would be nothing disappointing about such a result.
GPUs did not abolish CPUs.
They became spectacularly useful precisely because they were good at particular kinds of computation.
Modern AI systems depend upon heterogeneous architectures rather than a single magic processor.
A quantum processor becoming an important specialised component would still be an extraordinary technological achievement.
It simply would not satisfy the adolescent fantasy in which a silver refrigerator replaces the data centre and begins contemplating all possible universes before lunch.
That fantasy survives because “specialised accelerator” sounds insufficiently messianic.
Nobody wishes to stand upon a conference stage and announce:
Ladies and gentlemen, after thirty years of research we believe we may eventually have a useful coprocessor.
Even if that sentence proves worth hundreds of billions of dollars.
Technological culture has developed a terrible allergy to proportion.
Everything must change everything.
Every device must remake civilisation.
Every algorithm must disrupt an industry.
Every research programme must become a platform.
Every prototype must become an ecosystem.
One wonders how electricity survived its first decade without a venture-capital vocabulary.
Quantum computing possesses an additional rhetorical advantage: almost nobody outside the field wants to admit that he does not understand it.
This is very valuable.
If someone sells you a toaster, you may ask whether it makes toast.
If someone sells you a quantum strategy, you first ask what a qubit is.
The salesman has already escaped.
By the time everyone has discussed superposition, entanglement, interference, decoherence, error mitigation, surface codes and magic states, asking whether the company has saved twelve dollars on logistics seems positively uncultured.
Complexity has always been the finest tailoring nonsense can afford.
This does not mean that quantum computing is nonsense.
Quite the opposite.
It means the science is sufficiently profound that nonsense can stand near it without immediately being recognised.
The most effective technological hype never invents an entirely fictitious object.
It takes a real achievement and quietly removes the qualifications.
A researcher says, “Under these assumptions, for this workload, using this metric, we observed performance that exceeded the currently tested classical techniques.”
Publicity hears, “Quantum computing wins.”
A scientist says, “This experiment provides evidence of computation in a regime difficult for exact classical verification.”
A consultant hears, “Every Fortune 500 company needs a quantum transformation office.”
An engineer says, “We reduced initialization errors and increased circuit throughput.”
The newspaper hears, “The age of classical computers is ending.”
And somewhere the engineer turns back to the refrigerator.
One almost envies the refrigerator.
It is never required to write a thought-leadership article.
The physical complexity of these systems deserves attention because it dispels another popular illusion.
Quantum computing is sometimes described as though one merely replaces bits with qubits.
In reality, the machines require an elaborate environment of cooling, control, electronics, calibration, signal generation and classical processing.
IBM’s Quantum System Two description is wonderfully material: cryogenic infrastructure holds qubits near absolute zero, qubit-control electronics deliver signals to the QPUs, and classical runtime servers support hybrid workflows.
Quantum theory may be ethereal.
The plumbing is not.
The popular image of the modern superconducting quantum computer is the famous golden chandelier of cables and refrigeration stages descending inside a cryostat.
It looks magnificent.
It looks sufficiently expensive to persuade anyone nearby that history is occurring.
The actual quantum processor is considerably less theatrical.
Technology has always understood staging.
The first rule of selling the future is to ensure the future photographs well.
Yet none of this equipment is frivolous.
The extreme engineering exists because qubits are fragile.
Noise matters.
Control errors matter.
Readout errors matter.
Calibration matters.
Coherence matters.
State preparation matters.
Interactions with the surrounding environment matter.
The world has an infuriating tendency to interfere with quantum computers merely by existing.
Classical computers have spent decades becoming tolerant of ordinary life.
Quantum computers presently require something closer to diplomatic immunity.
Which brings us to fault tolerance.
This is where the distinction between physical and logical qubits becomes indispensable.
A physical qubit is a physical quantum system used to encode and manipulate information.
A logical qubit is encoded across physical resources in such a way that errors can be detected and corrected sufficiently well for reliable computation.
The practical promise of deep quantum algorithms depends on moving from delicate physical operations toward robust logical computation.
That transition is not an optional refinement.
It is the bridge between impressive quantum experiments and machines capable of sustaining very long calculations with controlled error.
IBM’s own roadmap makes this explicit by identifying Starling as its future fault-tolerant machine and setting out a path toward larger logical systems.
Again, there is nothing scandalous about not having arrived.
The scandal would lie only in pretending destination and departure are synonyms.
Aviation did not become fraudulent because the Wright Flyer was not a Boeing 747.
But it would have been mildly eccentric for the Wright brothers to sell transatlantic business class.
Quantum computing is currently in the peculiar position where primitive and advanced forms of the technology share the same convenient noun.
A five-qubit teaching device is a quantum computer.
A 120-qubit research processor is a quantum computer.
A future error-corrected machine with hundreds of logical qubits is a quantum computer.
A hypothetical mature architecture capable of economically transformative workloads is a quantum computer.
One word.
Several universes.
This semantic flexibility allows discussions to glide between what has been demonstrated and what remains aspirational.
The physical machine exists; therefore the future machine has “arrived.”
The future machine could transform chemistry; therefore today’s machine is “transformative.”
A small demonstration works; therefore scaling is treated as an engineering detail.
The noun has committed the argument before the evidence is allowed to speak.
IBM’s ten-year retrospective on cloud quantum computing provides a useful corrective because it documents genuine progress without needing to pretend completion. IBM notes that it put quantum hardware on the cloud in May 2016 and that its systems have developed from early small devices to processors with more than one hundred qubits and a much more mature software stack.
That is substantial.
Ten years is not an eternity in difficult physics.
The transformation of cloud-accessible quantum hardware from tiny experimental systems into the present ecosystem is significant.
Researchers have been able to develop software against real machines rather than abstractions.
Students have learned on genuine quantum processors.
A developer community has emerged.
Hardware, compilers and tooling have matured.
The correct response to such progress should be admiration.
Apparently admiration is no longer enough.
We require astonishment.
We require inevitability.
We require a civilisation-changing narrative before lunch.
Modern technological culture treats incremental progress as though acknowledging increments were an insult.
Yet all actual engineering consists of increments.
The transistor did not become less important because semiconductor technology took decades to mature.
The internet did not cease to matter because early networks were limited.
Aircraft were not failures because the first machines carried fewer passengers than Heathrow.
Quantum computing should be permitted the dignity of being difficult.
Its defenders sometimes do the field a disservice by demanding that every advance be interpreted as proof of an imminent revolution.
A scientific field should not need us to exaggerate it.
If it succeeds, reality will eventually become sufficiently rude that sceptics will be forced to concede.
That is the lovely thing about useful technology.
It starts working.
One can stop holding panels about its potential.
Nobody now holds conferences entitled Could Databases Transform Accounting?
The argument has become unnecessary.
Useful technologies eventually acquire the vulgar distinction of customers who require no ideological instruction.
The most interesting question is therefore not whether quantum computing is real.
That debate is silly.
The question is when, where and under what economic conditions it becomes superior.
Those conditions matter enormously.
Suppose a quantum calculation produces an answer faster than a particular classical algorithm but requires immense classical preprocessing, cryogenic infrastructure, calibration and post-processing.
What is the total system cost?
Suppose a quantum method offers an asymptotic advantage but the crossover occurs at a problem size no commercial user needs.
What is the practical value?
Suppose a quantum system obtains superior accuracy on a specialised scientific task but must operate inside a multimillion-dollar facility.
What is the competing classical price?
Suppose a speedup exists but classical hardware becomes cheaper faster than quantum hardware improves.
Which curve wins?
These are not hostile questions.
They are the questions one asks when technology leaves physics and enters economics.
Quantum computing will become genuinely disruptive only when somebody rationally purchases quantum computation because it is the best available means of obtaining an outcome, rather than because having quantum in the annual report looks strategic.
There is an enormous difference between research demand and economic demand.
Governments will fund frontier research because nations dislike being technologically surprised.
Corporations will fund experiments because competitors are funding experiments.
Universities will create programmes because students want quantum credentials.
Consultancies will recommend preparedness because preparedness produces consulting engagements.
All of these behaviours can be perfectly rational without proving that the underlying technology has already achieved widespread commercial superiority.
An ecosystem can grow around uncertainty.
Indeed, uncertainty is often excellent fertiliser.
A country cannot afford to ignore quantum computing because another country might achieve something important first.
A bank cannot entirely ignore it because cryptographic implications matter.
A pharmaceutical company may sensibly explore quantum chemistry in case the technology becomes useful.
A technology firm wants expertise before the market matures.
Thus investment grows partly because nobody can afford to be the only institution that ignored a potentially transformative field.
This is rational option value.
It is not proof of present utility.
The two are constantly confused.
The existence of a quantum division does not prove quantum advantage any more than possessing an innovation department proves innovation.
Corporate organisation charts are among the weakest instruments in experimental physics.
There is also a wonderfully self-reinforcing institutional cycle.
Research funding produces laboratories.
Laboratories produce demonstrations.
Demonstrations justify additional programmes.
Programmes train specialists.
Specialists require employers.
Employers create teams.
Teams require strategies.
Strategies create consultancy.
Consultancy produces market forecasts.
Market forecasts justify investment.
Investment demonstrates that “the market believes.”
At the end, the amount of money invested becomes evidence for the proposition that justified investing the money.
Finance has discovered perpetual motion.
The antidote remains measurement.
Not cynicism.
Measurement.
Show the workload.
Show the hardware.
Show the error.
Show the algorithm.
Show the best classical competitor.
Show the end-to-end runtime.
Show the preprocessing.
Show the post-processing.
Show the capital cost.
Show the operating cost.
Show the energy requirement.
Show the accuracy.
Show the scaling assumptions.
Show the economically relevant crossover.
Then we may talk about advantage.
This should not be controversial.
Yet technological enthusiasm often treats demands for measurement as hostility because measurement has a vulgar habit of reducing narratives to quantities.
A number cannot be inspirational merely because the keynote speaker lowers his voice.
Simulation deserves the same discipline.
IBM’s Qiskit ecosystem provides excellent tools for classical simulation of quantum circuits. These tools are valuable precisely because they permit developers to test algorithms, reproduce ideal circuits, introduce noise models and understand expected behaviour before using scarce or expensive quantum hardware.
This is excellent scientific practice.
But let us retain nouns.
A simulation of a quantum computer is a simulation.
A physical QPU is a physical QPU.
An error-corrected logical architecture is another thing again.
A commercially superior application is yet another.
There is no intellectual gain in merging them.
There is considerable marketing gain.
The difference resembles the distinction between a map and a journey.
A map of Paris may be extraordinarily accurate.
It may be interactive.
It may model traffic.
It may predict travel times.
It may incorporate every building in extraordinary detail.
It remains possible to own the finest map of Paris ever constructed while sitting in Birmingham.
Only the technology sector could announce this as mobility.
And yet the classical simulator deserves respect.
It is often treated as though it were merely a temporary crutch until quantum hardware becomes sufficiently powerful.
In reality, classical simulation and classical optimisation will probably remain essential parts of quantum development indefinitely.
Certain circuit families can be simulated efficiently.
Small systems remain easy to check.
Special structures permit clever classical treatment.
Noise modelling matters.
Verification matters.
The continuing classical challenge will force quantum researchers to demonstrate stronger results.
The classical computer is not merely the old order awaiting execution.
It is the adversary that keeps the quantum claim honest.
IBM’s Quantum Advantage Tracker embodies precisely this competitive relationship. It recognises that the frontier moves and that apparent quantum advantages must be continually tested against advancing classical methods.
This is possibly the healthiest development in the entire debate.
If a classical algorithm catches a quantum result, that is scientific progress.
If a quantum method survives the strongest classical attack, the evidence for advantage becomes better.
Either way, knowledge improves.
Only marketing requires one side to win permanently by press release.
There is a lovely irony here.
The stronger classical computing becomes, the more meaningful a genuine quantum victory becomes.
Every classical improvement raises the evidentiary bar.
Every successful quantum result above that bar becomes harder to dismiss.
The proper friend of quantum computing should therefore welcome aggressive classical competition.
Nothing purifies a technological claim like an enemy trying to kill it.
This is why the correct attitude toward quantum computing is neither breathless devotion nor theatrical dismissal.
It is disciplined impatience.
The machines are real.
The engineering is formidable.
The progress is measurable.
The potential applications are serious.
The fault-tolerant destination remains ahead.
The commercial scope remains to be proven workload by workload.
The classical competition remains alive.
One can hold all these propositions simultaneously without suffering cognitive injury.
What one cannot reasonably do is take the existence of the first four and treat the fifth and sixth as automatic consequences.
IBM may ultimately succeed spectacularly.
Starling may arrive.
Error correction may scale.
Useful logical qubits may become plentiful enough to sustain computations inaccessible to any economically comparable classical system.
Quantum chemistry may yield industrial advantages.
Materials modelling may improve.
Optimisation may find important niches.
New applications may emerge that nobody currently anticipates.
Nothing in scepticism requires failure.
Scepticism merely refuses to award the medal before the race.
If the decisive machines arrive, sarcasm will become obsolete.
This is the ideal fate of sarcasm.
A technology should eventually become so useful that mockery sounds silly.
The day a fault-tolerant quantum system solves commercially important problems more cheaply or more effectively than the best classical alternative, one need not debate whether quantum computing matters.
The invoice will settle the matter.
Until then, however, we are entitled to distinguish a prototype from an economy.
A simulator from the object simulated.
A physical qubit from a logical qubit.
A roadmap from a delivery note.
A benchmark from a business model.
An advantage candidate from an industry transformed.
These are not pedantic distinctions.
They are precisely the distinctions that prevent science becoming theatre.
IBM’s own documentation gives us enough honesty to make them.
It tells us where the real hardware is.
It tells us when we are running a simulator.
It tells us that hybrid classical-quantum computation is central to the architecture.
It tells us what its new processors can currently do.
It tells us that Starling is planned for 2029.
It tells us that larger fault-tolerant systems remain roadmap objectives.
It tells us that quantum advantage must be tested against moving classical alternatives.
One need not invent an accusation.
One need only read carefully.
That is perhaps the cruelest thing one can do to technological marketing.
Read the documentation.
Read the dates.
Read the qualifiers.
Read the difference between is and will.
Read the difference between a QPU and a simulator.
Read the distinction between physical and logical computation.
Read what the machine actually does.
The future usually becomes rather less immediate after one reaches the footnotes.
And perhaps rather more interesting.
Because once the theatre is removed, what remains is genuine engineering.
IBM has spent a decade putting increasingly capable quantum processors into the hands of users through the cloud.
It has developed substantial software infrastructure.
It has built machines such as Heron and Nighthawk.
It has increased circuit throughput.
It is pursuing error correction.
It is integrating quantum and classical computing.
It is publicly mapping a path toward fault tolerance.
Those achievements do not need a fairy tale.
Science becomes more impressive, not less, when one states accurately how difficult the unfinished work remains.
The mature position is therefore wonderfully unfashionable.
Quantum computing exists.
Useful quantum computing exists in limited forms.
Quantum advantage is becoming an empirical contest rather than merely a theoretical aspiration.
Large-scale fault-tolerant quantum computing has not yet become an ordinary commercial reality.
Classical computers remain extraordinarily competitive.
Classical simulators remain indispensable.
Hybrid systems are likely to matter enormously.
And nobody yet knows with certainty how broad the eventual economic domain of quantum computation will be.
This uncertainty is not embarrassing.
It is called research.
The embarrassment begins only when uncertainty puts on a suit and calls itself inevitability.
I rather hope IBM succeeds.
Not because the publicity deserves vindication.
Publicity deserves nothing.
But because the physics is extraordinary, the engineering challenge is real and a genuinely fault-tolerant quantum computer capable of solving valuable problems beyond practical classical reach would be one of the great technological accomplishments of the age.
When that machine exists, we should celebrate it without reservation.
Until then, the classical computer remains beside it.
Designing it.
Programming it.
Simulating it.
Controlling it.
Benchmarking it.
Correcting it.
Interpreting it.
And occasionally being asked to simulate the thing that has supposedly made classical computation obsolete.
There is something almost touching about that.
The condemned man is building the guillotine, writing the operating manual and running the rehearsal.
IBM’s quantum machine may indeed be the future.
But the future has always had one commercial disadvantage.
It is not presently available.
The simulator is.
And it runs, with exquisite irony, on the computer whose funeral we have been attending for years.
Perhaps one day the quantum computer will finally render the classical machine secondary.
On that day the classical computer might reasonably ask for a small plaque.
Not an extravagant one.
Just something tasteful:
HERE STOOD THE MACHINE THAT SIMULATED ITS SUCCESSOR UNTIL ITS SUCCESSOR COULD ACTUALLY DO THE JOB.
It would be the first honest monument in Silicon Valley.
And, unlike the roadmap, one might even be able to deliver it on time.
References
Davis, R. (2026, May 4). A decade of quantum on the cloud. IBM Quantum Computing Blog.
Gambetta, J., & Davis, R. (2026, February 23). Quantum Advantage Tracker: The race to advantage. IBM Quantum Computing Blog.
Haas, H., McKay, D., & Davis, R. (2026, August 31). IBM Quantum Nighthawk r2—more circuits, faster. IBM Quantum Computing Blog.
IBM. (2026). Build and run your first quantum program. IBM Quantum Learning.
IBM. (2026). Exact and noisy simulation with Qiskit Aer primitives. IBM Quantum Documentation.
IBM. (2026). IBM Quantum computing: Hardware and roadmap. IBM.
IBM. (2026). Local testing mode. IBM Quantum Documentation.
IBM. (2026). Migrate to local simulators. IBM Quantum Documentation.
IBM. (2026). Quantum Roadmap. IBM Technology Atlas.
IBM. (2026). Quantum 2030. IBM Technology Atlas.
Kandala, A., Javadi-Abhari, A., & Gambetta, J. (2026, July 30). Quantum advantage through trusted quantum computation. IBM Quantum Computing Blog.