The Error is the Point

Quantum computers have always promised extraordinary things. They have also always broken. The errors are not a bug to be patched — they are a consequence of the physics.

May 24, 2026

Quantum computers have always promised extraordinary things. They have also always broken — not because of poor engineering, but because of the physics. The qubits in a quantum computer are extraordinarily fragile. They interact with their environment, lose their quantum state, and make errors at rates that would render any classical computer unusable. The field has known this since the beginning. It has also known what is required to fix it.

In December 2024, Google announced that its Willow quantum chip had crossed a threshold that quantum computing researchers had been working toward since the mid-1990s. Understanding why that matters — and why it is still only the beginning — requires a short account of what quantum computing actually is, and why errors are so central to the challenge.

Why errors are everything

A classical computer stores information as bits: each one is definitively a zero or a one. A quantum computer stores information as qubits, which can exist in what is called a superposition of zero and one simultaneously. This is not a metaphor — it is the actual quantum mechanical state of the physical system. What makes quantum computers potentially powerful is that a system of qubits can represent and process an exponentially large number of possible states at the same time, performing certain kinds of calculations far more efficiently than any classical approach.

The catch is that this quantum state is extraordinarily delicate. The moment a qubit interacts with its environment — with vibration, heat, stray electromagnetic radiation, even the act of measurement — the superposition collapses. The qubit behaves like a classical bit. The quantum advantage disappears.

This process is called decoherence. Managing it is the central challenge of quantum computing.

The solution theorists developed, quantum error correction, sounds counterintuitive at first. Instead of trying to build qubits that don’t make errors (which turns out to be essentially impossible with current technology), you build a logical qubit from many physical qubits, encoding the information redundantly and using the correlations between the physical qubits to detect and correct errors before they propagate. A logical qubit might require tens, hundreds, or eventually thousands of physical qubits to maintain reliably.

This creates a daunting scaling problem. If each logical qubit requires a hundred physical qubits, and a useful quantum algorithm requires a thousand logical qubits, you need a hundred thousand physical qubits all operating with error rates low enough for the correction scheme to work. For most of quantum computing’s history, this seemed so distant as to be almost theoretical. Then, in late 2024, Google published something that changed the conversation.

The error is not a bug to be patched. It is a consequence of the physics. Quantum error correction doesn’t prevent errors — it detects and corrects them before they matter, using redundancy across many physical qubits.

Google’s Willow chip demonstrated, for the first time in any quantum computing platform, that making a quantum error-correcting code larger actually reduced the error rate of the encoded logical qubit, and that this improvement was exponential. This is the property theorists had predicted was necessary for error-corrected quantum computing to work at scale, and it had never been conclusively shown experimentally. The larger the surface code, the more physical qubits devoted to error correction, the better the logical qubit performed.

The scaling law held. The path, at least in principle, was clear.

Willow’s 105 physical qubits also completed a specific benchmark calculation in roughly five minutes that would require a classical supercomputer an estimated 10 to the power of 25 years to perform. That number is so large as to be meaningless in everyday terms — the age of the universe is around 14 billion years, or roughly 10 to the power of 10. The point is not the absolute comparison. It is the direction and magnitude of the gap.

Three roads, one destination

The quantum computing field in 2025 and 2026 is characterised by genuine competition between categorically different technological approaches — not just different companies building the same kind of machine, but different physical architectures with different advantages, different error profiles and different paths to scale. Understanding them does not require a physics degree. It requires grasping one central distinction: whether error correction is done primarily in software, or baked into the hardware.

Google and IBM have both built their systems around superconducting qubits — tiny circuits, cooled to near absolute zero, that exhibit quantum behaviour. IBM’s roadmap is the most detailed in the industry. Its Nighthawk processor, released in late 2025, delivered 120 high-quality physical qubits and introduced couplers that extend connectivity across the chip. IBM targets verified quantum advantage, solving a problem better than any classical method, by the end of 2026, and a full fault-tolerant machine with 200 logical qubits by 2029. Its approach emphasises steady incremental progress and a clear engineering roadmap.

IonQ and Quantinuum build their qubits from individual ions — electrically charged atoms trapped in electromagnetic fields. Trapped-ion systems have better error rates per operation than superconducting systems, and any two trapped-ion qubits can interact directly without the connectivity constraints of a chip. The trade-off is speed: trapped-ion operations are slower. QuEra and Atom Computing use neutral atoms, manipulated with precisely aimed laser beams, which offer their own connectivity advantages. IEEE Spectrum’s Top Tech 2026 report named neutral-atom systems among the most promising near-term candidates for error-corrected demonstrations.

The quantum computing landscape in 2026:

Google (Willow, superconducting) — Demonstrated below-threshold error correction Dec 2024
IBM (Nighthawk/Kookaburra, superconducting) — Quantum advantage target: end 2026; fault-tolerant: 2029
Microsoft (Majorana 1, topological) — Qubit intrinsically resistant to errors; claims contested by peer reviewers IonQ / Quantinuum (trapped ions) — Lower error rates per gate; slower clock speed QuEra / Atom Computing (neutral atoms) — Strong connectivity; early error-corrected demonstrations
Market size 2025: ~$1.8–3.5bn
Projected 2030: up to $20bn
Quantinuum private valuation: $10bn PsiQuantum: $7bn SandboxAQ: $5.75bn
One million qubits = widely cited threshold for transformative quantum computing

Microsoft’s approach is categorically different from all of the above. In February 2025, the company announced Majorana 1, a chip built around what it calls topological qubits — a form of qubit that encodes information in the global geometric properties of a physical system rather than in the state of individual particles. The theoretical appeal is significant: a topological qubit should be inherently resistant to local errors because its quantum information is distributed, not localised. Braid the Majorana particles around each other and you have performed a computation; local noise cannot accidentally unbraid them.

The announcement attracted intense scrutiny. The Nature paper came with an editorial note stating that the results did not constitute evidence for topological modes as claimed in Microsoft’s press release. At the American Physical Society’s global summit in March 2025, physicists including Henry Legg of St Andrews and Eun-Ah Kim of Cornell went on record with doubts about Microsoft’s verification methods — the field’s independent assessment concluding that the evidence fell short of establishing topological modes, and that the topological approach remained the furthest from practical demonstration of the competing architectures. Microsoft’s researchers disputed this forcefully. Scott Aaronson, a prominent quantum computing theorist at the University of Texas, captured the independent position precisely: the approach is worth pursuing, but the evidence so far is partial and further independent replication is required.

By mid-2025, Microsoft had published additional work demonstrating a tetron qubit device, the next step in its roadmap, showing that the error protection mechanism behaved as theoretically predicted. The picture was clearer. The independent verification was not yet there.

What becomes possible

The question of what a working quantum computer could actually do is easier to answer than the question of when one will exist.

Drug discovery and materials science. A fault-tolerant quantum computer could simulate chemical reactions with a precision no classical supercomputer will ever achieve, regardless of how much faster they become. The implications for discovering new medicines, designing new catalysts, modelling protein folding are real. A simulation that shows exactly how a drug candidate binds to a protein in living biology is not a fantasy. It is the purpose for which quantum computing was originally conceived.

Cryptography. Most of the encryption protecting the internet — the secure connections used for banking, e-commerce, and private communication — relies on the computational difficulty of factoring very large numbers. A sufficiently capable quantum computer running Shor’s algorithm could break RSA encryption. This is not a near-term threat: the quantum computers that exist today are nowhere near the scale required.

But a 2025 analysis by Google Quantum AI revised earlier estimates of around twenty million physical qubits down to approximately one million. As error correction improves, the timeline for this threat becomes less abstract. The transition to post-quantum cryptographic standards — already underway, with NIST finalising new standards in August 2024 — is not an abundance of caution. It is a necessary precaution against a threat that will arrive eventually.

Optimisation. Many of the most economically important computational problems — logistics routing, financial portfolio optimisation, supply chain management, climate modelling — are optimisation problems, in which you are searching for the best solution across an enormous space of possibilities. Quantum approaches to these problems may deliver advantages that are harder to characterise than the clear theoretical guarantees of drug simulation or cryptography, but that could be commercially significant. IBM, working with partners including Boeing, Cleveland Clinic and Oak Ridge National Laboratory, is actively investigating where quantum advantage first becomes tangible in real industrial problems.

A fault-tolerant quantum computer could simulate chemical reactions at a level no classical machine will ever match. Drug discovery, materials science, and cryptography are the three applications where the stakes are clearest.

The honest position on timelines is the one offered by IEEE Spectrum’s Top Tech 2026 report: we will not get there in 2026. The quantum computers that exist today, including Google’s Willow and IBM’s Nighthawk, are capable of impressive demonstrations. They are not capable of the sustained, reliable, large-scale error-corrected computation that drug discovery or cryptographic breaking would require. IBM targets fault-tolerant machines by 2029. The same report puts the first practical, large-scale quantum applications in the early 2030s at the optimistic end — and considerably later at the realistic end.

Jensen Huang’s January 2025 estimate of fifteen to thirty years drew enough pushback that he publicly walked it back two months later; Bill Gates suggested the possibility of three to five. The range of expert opinion remains wide.

What has changed, what makes 2025 and 2026 different from the decade before them, is that the theoretical framework has been experimentally confirmed. The below-threshold demonstration matters not because it delivers useful computation today, but because it validates the entire foundation that error-corrected quantum computing depends on. The scaling law holds. The path is real. The question has shifted from whether this can work to when.

Who this is actually for

The case for quantum computing is most often made in terms of what it makes possible — better drugs, stronger materials, faster computation. The more important question is who it is being developed for. The molecular simulations quantum computers could enable would change what research chemists can do in ways classical computers structurally cannot. The relevant question is whether the technology is being developed with those intended beneficiaries clearly in view — the researchers, the patients who might receive better drugs, the communities whose data security depends on cryptographic standards. Or whether the primary beneficiaries turn out to be the companies and investors who will control access to it.

The access question is not resolved by Quantum-as-a-Service platforms existing. IBM, Google and Microsoft all offer cloud-based quantum access to researchers worldwide. Whether that access is available to the researcher in Singapore, in Nairobi, in São Paulo who needs quantum capabilities for their scientific work — whether they can afford it, and use it — is not a technical question. It is an economic and political one.

The difference between quantum capability being broadly accessible and quantum capability being concentrated in the hands of governments and large corporations is not determined by the physics. It is determined by pricing, by infrastructure, and by decisions that technology companies and policymakers are making now.

The most urgent question is specific: the RSA cryptography threat. Quantum computers will eventually be able to break the encryption that currently protects much of the world’s digital infrastructure. The timeline is not known precisely, but it is not infinite. Google has set a 2029 internal deadline for completing its own migration to post-quantum cryptography — well ahead of NIST’s 2035 guideline and NSA’s 2031 target. The organisations that are actively planning for this transition — implementing the NIST standards, auditing their systems, migrating to quantum-resistant encryption — are acting on the right information.

Those treating this as a distant concern are making a mistake that is difficult to correct retroactively: data that is encrypted today but vulnerable to future quantum attack is being harvested now by adversaries planning to decrypt it later. This is the harvest now, decrypt later threat, and it is operational, not theoretical.

The decisions that can’t wait

Google’s Willow demonstration crossed a threshold that quantum computing theorists had worked toward since the mid-1990s. How much does that milestone matter to you — does it feel like a genuine turning point, or another piece of hype in a field that has promised results for decades?

The encryption protecting your banking, your communications and your medical records is vulnerable to a sufficiently capable quantum computer. Post-quantum cryptography standards have been finalised. Does the organisation you work for, or depend on, have a plan for migrating to them? If not, why not?

Drug discovery, materials science, logistics — the applications that stand to benefit most from quantum computing are spread across industries and geographies. Who controls access to the quantum computers that make those benefits possible? Does it matter?

Microsoft’s topological qubit claims generated genuine scientific disagreement — between company researchers and independent physicists, in peer-reviewed journals and at conferences. How do you weigh corporate announcements about breakthrough technologies against the more cautious assessments of independent scientists? What standard of evidence would you need to believe this technology had arrived?

Quantum computing has been described as revolutionary for so long that the word has lost some of its force. It has been genuinely difficult, the promises have repeatedly outrun the delivery. That is the right response to the actual physics rather than a failure of ambition. The wait is now measured in years, not decades.

My opinion

The claim that quantum computing has the potential to change everything has been made before — about electricity, about computing, about the internet. Sometimes it was right.

The challenge here is not building the machine. It is understanding what the machine is doing. If the computation happens inside error-correcting codes that no individual can read back, how do we know when something is wrong? If a company uses quantum simulation to discover a drug, how does a regulator verify the reasoning?

Trust requires interpretability. Quantum computing, at scale, may not offer it in a form that governance can work with.

When the technology becomes commercially viable, and the evidence now suggests that it will, companies will need to test it, govern it, audit it, and account for what it decides. The questions of how to use it, how to verify it, how to control it and who decides are not technical questions. They are political and regulatory ones. The infrastructure to answer them does not yet exist. The physics has been ahead of the governance for a generation. The gap is still widening.

The Next Evolution

Discussion about this post

Ready for more?