Technology · Future technologies
Quantum Computing: The Era of Logical Qubits Begins
Quantum computers were seen as hopelessly error-prone – now the curve is turning: error correction finally scales, and logical qubits grow measurably.
By Boaz Lichtenstein
Quantum computing’s curse was always the same: its computing units – qubits – are so sensitive that any disturbance wrecks the calculation, and the more of them you built in, the more errors you got. Critics saw that as a dead end with no way out. That exact curve has recently turned – which is why the current phase is the most interesting since the field was invented.
Key takeaways
- The central breakthrough is exponential error suppression: more hardware now measurably means more reliability instead of more errors.
- Logical qubits – error-corrected units built from many physical qubits – are the metric that matters, not the physical qubit count alone.
- Current records sit in the double-to-triple-digit range of logical qubits; a cryptographically relevant quantum computer would need orders of magnitude more.
- Practical value is emerging first in niches with a genuine quantum fit: chemistry, materials research, complex optimisation – not general-purpose computing.
- For businesses, the most urgent consequence today isn’t access to quantum hardware, but migrating to post-quantum cryptography.
The turning point: error correction that scales
Why was error correction seen as the field’s unsolved problem for decades? Because more qubits used to mean more sources of error too – a vicious circle that immediately ate up any gain in scale. The breakthrough has a technical name: exponential error suppression. Google’s Willow processor showed that a logical qubit’s error rate markedly falls rather than rises with every increase in the size of the correction code – proof that more hardware finally means more reliability. Since then, logical-qubit records have been tumbling: QuEra demonstrated 96 of them in early 2026 and is targeting commercial systems with around 100 logical qubits; in parallel, real-time decoders are maturing that correct errors during the calculation instead of analysing them afterwards. Germany is in the game too – from systems in Garching to Infineon’s ion-trap chips. Translated: the field has moved from the “whether” of error correction to the “how fast can we scale it” – a change of category, not an increment.
Physical and logical qubits: why the headline number alone says nothing
Why doesn’t a “5,000 qubits” headline work as a measure of progress? Because physical qubits – the actual hardware components – are individually extremely error-prone: even a stray cosmic ray or a tiny temperature fluctuation can destroy their state. Only once you bundle dozens to hundreds of physical qubits into a logical qubit that continuously detects and corrects errors do you get a unit you can actually rely on for computation. A system with many physical but few logical qubits is like an engine with impressive horsepower on paper that stalls after a few seconds – the horsepower figure impresses, but what counts is how the car actually drives. That’s precisely why the field has settled on measuring progress in logical rather than physical qubits.
From experience: anyone reading news about quantum computers should first check whether a report is talking about physical or logical qubits – the two figures are routinely conflated in coverage. A headline citing a high physical qubit count with no mention of error rate is a warning sign for marketing over substance; credible announcements now almost always state the number of logical qubits alongside the error rate achieved.
What quantum computers may become – and what they won’t
Will quantum computers one day replace ordinary PCs? No – the biggest misconception first: quantum computers are not faster general-purpose machines – for email, databases and streaming they remain irrelevant, often even slower and more expensive than classical hardware. Their strength lies in problems whose space of possibilities explodes classically: chemistry and materials research (simulating molecules rather than approximating them – from catalysts to battery chemistry, the bridge to our sodium-ion battery article), optimisation in logistics and finance, and – as the downside – cryptography: a sufficiently large quantum computer would break today’s public-key methods, which is why the post-quantum migration is already under way (see FAQ). A realistic timeline for the first commercial value: niches by the end of this decade, relevant breadth in the 2030s – with the usual caveat that manufacturer roadmaps price in optimism.
Quantum computer versus classical supercomputer, compared
| Criterion | Classical supercomputer | Quantum computer (current state) |
|---|---|---|
| Basic unit | bit (0 or 1) | qubit (superposition state) |
| Strength | all everyday tasks, reliable | very specific problem classes |
| Maturity | decades of industrial use | research and early stage |
| Error-proneness | minimal, well controlled | high, error correction a central focus |
| Access | standard cloud, available everywhere | specialist cloud offerings, few providers |
Who’s driving the development right now
Who’s building the systems making headlines right now? The field is broader than it’s often portrayed: Google and IBM are pursuing superconducting chip architectures, QuEra and other providers are betting on neutral atoms, IonQ and Infineon partners on ion traps, PsiQuantum on photonic approaches. Each architecture has its own trade-offs in error correction, scalability and operating temperature – no single “winner” has emerged yet. Germany and Europe are positioning themselves through research institutions and targeted funding programmes as a hub for fundamental research and ion-trap technology, rather than for the largest record-breaking systems – a realistic but relevant role in the global race.
It’s also notable how differently the timelines play out depending on architecture: superconducting chips are currently furthest ahead on error correction but need extreme cooling near absolute zero. Ion traps are considered especially precise but run at a slower clock rate. Photonic approaches promise room-temperature operation but are still at an early stage on error correction. For observers, that means: a single technology-path “winner” is by no means settled – it’s more realistic that different architectures will persist in parallel for different use cases, much as CPUs, GPUs and specialised chips have coexisted in classical hardware.
The most common misconceptions
- “More qubits is always better.” Wrong – without error correction, the raw qubit count is nearly meaningless; what matters is logical qubits and computational depth.
- “Quantum computers are just faster computers.” Wrong – they’re not in the running for the vast majority of everyday tasks, but are specialised tools for narrowly defined problems.
- “Q-Day is right around the corner.” Overstated – credible estimates put cryptographically relevant systems years away, even though preparation should already be starting now.
- “Quantum computing is pure science fiction with no practical relevance.” Also wrong – early pilot projects in chemistry, pharma and logistics are already running via cloud access to real quantum hardware.
- “A quantum computer runs through every possible answer to a problem in parallel simultaneously.” A popular simplification that distorts the actual mechanism: superposition enables certain computational tricks, but it’s no free pass for arbitrary speed-ups – only carefully constructed algorithms for suitable problem classes actually benefit.
The observer’s metric
As with the nuclear fusion reality check, the rule holds: you don’t measure progress by press releases, but by a single figure – error-corrected logical qubits multiplied by computational depth. If it keeps climbing at the current pace, the lab marvel becomes, within a foreseeable timeframe, a tool for specialist problems with genuine world-market relevance. If it stalls, it was just another leg in computing’s longest marathon. Anyone who wants to follow the debate without being dazzled by individual record announcements should apply the same sober metric used for other future technologies with long time horizons, such as in our nuclear fusion reality check.
The bottom line
For the first time, the error-correction curve is measurably pointing in the right direction – that’s the real difference from decades of past quantum-computing promises. Even so, the timeline stays long: relevant breadth of impact is a matter for the 2030s, not the next two or three years. For businesses, that means concretely: plan for post-quantum cryptography now, pursue pilot projects only where there’s a genuine subject-matter fit – and calibrate your own expectations against the observer’s metric rather than the headlines. Anyone who handles sensitive data day to day can already start asking, within their own IT landscape, which cryptographic building blocks still rely on the common, quantum-vulnerable methods – the shift to post-quantum standards is a multi-year project that should start early, not only once the “Q-Day” headlines get louder.