Ions and neutral atoms
Lesson 16 of 23 in the course “Quantum”. Free, with animations and a quiz.
A qubit made by nature
The second big approach: take a real atom. Strip one electron off and it becomes a charged ion, which an electric field can hold in a vacuum like an invisible cradle. Two energy levels inside the ion are 0 and 1, and gates are made with precise laser pulses. The trump card: all ions of one element are identical by nature. Superconducting qubits each differ slightly due to manufacturing; ions have no defects.
Helios: accuracy records
The leader of the ion approach is Quantinuum. In November 2025 it launched Helios: 98 barium ions, and since ions can be moved, any qubit can link with any other. Two-qubit gate fidelity is 99.921%. It hosted 48 error-corrected logical qubits. And in October 2026 the company reported a logical operation fidelity of 99.9995%, 10 times better than its own record. Next machines on the plan: Sol in 2027 and Apollo in 2029.
Neutral atoms: thousands of qubits
The third approach: ordinary neutral atoms, such as rubidium or caesium. Each is held by a focused laser beam called an optical tweezer. The atoms are arranged in a grid like pixels on a screen, which makes thousands of qubits easy to get. In 2025 Caltech built an array of 6,100 atoms. A Harvard, MIT and QuEra team showed a system of over 3,000 qubits that ran for more than two hours straight: lost atoms were refilled on the fly, up to 300,000 per second.
The atoms’ superpower: they can move
Laser tweezers can be moved, so atoms “drive up” to the neighbour they need right during the computation. For error correction that’s a huge plus: connections can be rearranged to fit the code. In March 2026 a Caltech team estimated that Shor’s algorithm on such atoms might need only 10,000–20,000 qubits. It’s a calculation, not a finished machine, but the industry reacted: that same year Google set up its own neutral-atom team, and Microsoft is building a machine with Atom Computing.
Who’s stronger at what
To simplify. Superconductors are the fastest, with gates in nanoseconds, but the qubits are noisier and connect only to neighbours. Ions are the most accurate and connect all-to-all, but gates are slower, and you can’t fit many ions in one trap. Atoms are the most scalable, easy to gather by the thousands and rearrange, but gates are slower too, and atoms sometimes get lost. No approach has won yet, so big players bet on several at once.
Check yourself
- What makes ion qubits fundamentally better than superconducting ones?
Correct answer: All ions are identical, and they connect all-to-all. Nature makes ions identical, and moving them gives links between any qubits. That’s where the accuracy records come from. - Why does moving atoms matter for error correction?
Correct answer: Connections can be rearranged to fit the code. Efficient codes need links beyond neighbours. Mobile atoms give that flexibility, which is why Shor qubit estimates for them are so low. - Which approach will win?
Correct answer: Unknown yet; each has its own strengths. Speed, accuracy and scale aren’t combined in one approach yet. So Google, Microsoft and others bet on several at once.
← Superconductors: Google and IBM Photons, Majorana and silicon →