Quantum computing: where the field actually stands

Quantum computing carries five metrics and 27 milestones here, and it is the field where the headline number is most often the wrong one.

The raw physical-qubit record is 6,100, set by Caltech in September 2025 with cesium atoms in optical tweezers — roughly ten times prior arrays. That number says almost nothing on its own. IBM's Condor reached 1,121 superconducting qubits back in 2023, and IonQ sold its first 256-qubit chip-based system to the University of Cambridge in May 2026. Physical count is a capacity figure, not a capability one.

The number that matters is logical qubits — error-corrected qubits that survive long enough to do work. That record is 96, set by QuEra and Harvard in January 2026 on a fault-tolerant architecture. Quantinuum runs up to 48 to 50. Gate quality underpins all of it: the threshold for practical error correction sits near 99.9%, and IonQ passed 99.99% in October 2025, with Quantinuum's Helios at 99.92% across 98 trapped-ion qubits.

What changed in 2026 is that error correction became routine rather than exotic. Microsoft and Quantinuum published in Nature a logical error rate up to 800× below the underlying physical qubits — the largest physical-to-logical gap independently validated. QuEra with Harvard and MIT showed qLDPC codes encoding 1,156 logical qubits into 2,304 physical ones, roughly 2:1, against the hundreds-to-one typical today. Atom Computing demonstrated sustained multi-round correction on neutral atoms, a first for that platform. Google's below-threshold Willow result in December 2024 was the turn; 2026 was the field acting on it.

Capital and states followed. Quantinuum priced the industry's first mega-IPO in June 2026, raising $1.68B at about $15.7B. IBM committed more than $10B over five years and is spinning up Anderon, a $2B superconducting-qubit wafer fab in Albany, then bought HRL Laboratories in August to add silicon-spin qubits to a roadmap built entirely on transmons. PsiQuantum broke ground near Brisbane on a utility-scale photonic machine aiming at a million physical qubits. France, Italy and Canada each made state purchases or commitments. Annual investment into the sector runs about $12.6B.

The gap to the goal is still enormous and worth stating plainly. A cryptographically relevant quantum computer — one that breaks RSA-2048 — needs millions of physical and thousands of logical qubits. The field is at 96 logical. That milestone sits in the 2030s on this tracker and is marked locked, not in progress. IBM's 2026 quantum-advantage claims are recorded as claims: company-led demonstrations paired with verification methods, on a question the field is still pressure-testing. Every figure links to its source on the metric pages.