A different kind of bit
A conventional computer stores information as bits, each either 0 or 1. A quantum computer uses quantum bits, or qubits, which are physical systems such as the spin of an electron or the state of a tiny superconducting circuit. Governed by quantum mechanics, a qubit can be in a combination of states, and several qubits can be linked through entanglement. Carefully designed operations can make wrong answers cancel out and right ones reinforce. This does not mean a quantum computer simply tries every answer at once, a common misconception; the gains come only for specific kinds of problems.
The 1980s proposals
The first ideas appeared in the early 1980s. Paul Benioff described in 1980 how a computer might in principle operate according to quantum mechanics. In 1981, at a conference on physics and computation, Richard Feynman argued that simulating quantum systems on ordinary computers becomes extremely demanding as they grow, and suggested that a machine built from quantum components might do it more naturally. Around the same time, the Russian mathematician Yuri Manin made related remarks.
In 1985 David Deutsch described a universal quantum computer, giving a formal model comparable to Turing’s model for ordinary computers, and showed a small problem where a quantum approach had an advantage over any classical method. The problem was artificial, but it showed that the idea was not empty.
Algorithms raise the stakes
Interest grew sharply in 1994, when Peter Shor published an algorithm for factoring large numbers. Much of the security of widely used public-key encryption rests on the assumed difficulty of factoring, so a large enough quantum computer running Shor’s method could in principle weaken those systems. In 1996 Lov Grover described a search algorithm that offers a more modest speedup for searching unsorted data. These results turned the subject from a curiosity into a topic for national security agencies and for cybersecurity researchers, who have since worked on new encryption methods intended to resist quantum attacks.
Early small machines
Building hardware proved much harder than writing equations, because qubits are fragile. Stray heat, vibration or radiation can disturb them, a problem known as decoherence. In 1995 Peter Shor and Andrew Steane independently proposed quantum error correction, which suggested that errors might be controlled using extra qubits, though at a large cost in numbers.
Late in the 1990s, groups using nuclear magnetic resonance on molecules in liquid demonstrated two-qubit and slightly larger computations. In 2001 an IBM team reported using a seven-qubit molecular system to factor the number 15, which is 3 times 5. It was a landmark as a demonstration, but the number is trivially small, and later analysis questioned how much quantumness such liquid-state systems really used.
Other approaches followed: trapped ions held in electric fields, superconducting circuits cooled close to absolute zero, photons and, more recently, neutral atoms. Each has strengths and weaknesses, and the field has not settled on a single winner.
Recent claims
In 2019 Google reported that its superconducting processor carried out a specific sampling task faster than the best known classical method would, calling this quantum supremacy. Other researchers argued that improved classical methods could do the task much faster than first estimated, and the disagreement shows how contested such milestones can be. Companies and universities now offer cloud computing access to small quantum processors, letting researchers experiment, much as time-sharing once opened early machines to wider use.
What remains unproven
Several large questions are still open:
- Whether machines with enough reliable, error-corrected qubits to run Shor’s algorithm on cryptographically useful numbers can be built, and when. Estimates of the number of physical qubits needed are very large, and timelines are uncertain.
- Which practical problems, beyond simulating molecules and materials, will actually benefit. For many everyday tasks, quantum computers are not expected to help.
- Whether the engineering can scale without costs growing too quickly.
Current devices are often described as noisy and intermediate in scale: real, but too small and error-prone for most useful calculations. The honest summary is that the theory is well established, small experimental machines exist, and the commercially transformative version has not yet been shown to work. Predictions from enthusiasts and skeptics alike should be treated with caution, a point familiar from the history of artificial intelligence.