For years, quantum computing headlines were a numbers game: 50 qubits, 100 qubits, 1,000 qubits. But raw qubit counts were always misleading, because those qubits were too error-prone to run anything useful for long. The real story of 2025–2026 is quieter and far more important: quantum computers are learning to correct their own errors.
The problem: qubits are fragile
A normal computer bit is either 0 or 1, and it stays that way. A qubit can hold a blend of both at once (superposition) and link with other qubits (entanglement). That’s where quantum computing’s power comes from.
The catch is that qubit states are incredibly delicate. Stray heat, electromagnetic noise or vibration can corrupt them in a fraction of a second. On early machines, errors piled up so fast that calculations turned to noise before they finished.
The fix: logical qubits
The solution is quantum error correction. Instead of trusting one physical qubit, you spread information across many physical qubits that together form one logical qubit. Extra “helper” qubits constantly check for errors and fix them without disturbing the actual calculation.
The crucial test is whether this gets better as you scale. If adding more physical qubits added more errors than it fixed, error correction would be a dead end.
2026: the numbers are moving
- Below threshold. Google’s Willow chip showed that each step up in error-correcting code size cut the logical error rate by roughly half. Bigger really is better, which is the result the field had been waiting decades for.
- 96 logical qubits. In January 2026, QuEra reported 96 logical qubits encoded from 448 neutral atoms in Nature, using efficient “high-rate” codes. That’s the largest verified logical qubit count so far.
- Others close behind. Quantinuum, Microsoft with Atom Computing, and IBM are all publishing steady logical-qubit progress on different hardware.
- Clear roadmaps. IBM aims for Starling in 2029: a fault-tolerant machine with 200 logical qubits running 100 million operations. Many researchers see around 100 stable logical qubits as the next big milestone.
What will quantum computers actually be good for?
A quantum computer won’t replace your laptop, speed up your web browsing or run your games. It’s a specialist tool for problems that overwhelm classical computers:
- Chemistry and materials: simulating molecules to design new drugs, batteries, catalysts and fertilisers
- Physics: modelling materials like superconductors
- Optimisation: some logistics, scheduling and finance problems
- Cryptography: a large enough fault-tolerant quantum computer could break much of today’s public-key encryption
That last point is why governments and tech companies are already acting, years before such a machine exists. And that’s the subject of the final post in this series.
Hype check
Be sceptical of claims that quantum computers are “outperforming” classical ones on useful problems today. Most such demonstrations are carefully chosen tasks with no practical value. The honest picture: error correction works, it scales, and useful machines are plausibly a few years away, not decades. That’s still a big deal.
Next in the series: Harvest Now, Decrypt Later: The Post-Quantum Deadline Is Here
Sources: Quantum Zeitgeist: Logical qubit leaderboard 2026 · Google Quantum AI: error-corrected quantum computers (arXiv) · IBM: Path to fault-tolerant quantum computing · The Quantum Insider
