Quantum computing is surrounded by two myths: that these machines are simply faster versions of ordinary computers, and that they try every possible answer at once. Neither is right. A quantum computer is a different kind of machine, expected to be dramatically better at a narrow class of problems and no help at all for most everyday tasks.
What a qubit is
An ordinary computer stores information as bits, each either 0 or 1. A quantum computer uses qubits, which can be placed in a combination of 0 and 1, a property called superposition, and which can be linked so that their states depend on each other, called entanglement. A quantum algorithm is designed so that the paths leading to wrong answers cancel each other out while the path to the right answer is reinforced. The speedup comes from that careful choreography, not from brute force.
Qubits can be built in several ways: from superconducting circuits cooled close to absolute zero, from individual charged atoms held by electric fields, from neutral atoms held by lasers, or from particles of light. No approach has clearly won.
The error problem
Qubits are fragile. Heat, vibration and stray electromagnetic fields disturb them, and every operation introduces errors. Today’s machines are limited less by how many qubits they have than by how often those qubits make mistakes.
The long-term remedy is quantum error correction, which spreads information across many physical qubits to create a smaller number of reliable logical ones. The catch is that adding more qubits to a code only helps if each additional layer reduces errors rather than adding more of them, a threshold that Google says had been an open challenge since error correction was proposed in 1995. In a paper in Nature in December 2024, Google reported crossing it on its Willow chip. Its largest code, using 101 qubits, cut the logical error rate by a factor of 2.14 each time the code grew in size by two steps, and the encoded memory outlasted the best physical qubit on the chip by a factor of 2.4. It was a landmark for the field, though it was a memory experiment, not a working computation.
What the companies are promising
IBM’s public roadmap targets a machine it calls Starling by 2029, which it describes as a modular, error-corrected system running 100 million operations on 200 logical qubits, followed by a larger system, Blue Jay, in 2033 or later. These are targets, not delivered systems, and IBM notes that its timelines may change.
The difference between physical and logical qubits is crucial when reading such figures. A machine with thousands of physical qubits may offer only a handful of reliable logical ones.
What “advantage” means so far
Quantum advantage means doing something a classical computer cannot do in reasonable time. In October 2025, Google reported in Nature a measurement it called verifiable, an algorithm named Quantum Echoes run on 103 qubits. Google estimates the task would take a leading supercomputer about 13,000 times longer, a figure that is the company’s own estimate. Such claims are routinely tested by classical-computing researchers who look for faster conventional methods.
Even Google’s own framework for the field is candid about the state of play: no end-to-end quantum application has yet been implemented in hardware with a conclusive advantage on a problem of real-world consequence. The experiments so far are demonstrations, chosen because they are hard for classical machines, not because anyone needs the answer.
Where quantum could eventually help
The leading candidates are simulating molecules and materials, since nature itself is quantum, and certain optimization and sampling problems where the benefit remains uncertain. Quantum computers are not expected to speed up web browsing, word processing or video, and some hard problems may stay hard for them too.
The encryption connection
The application that gets the most attention is breaking encryption. A known quantum algorithm could, on a large enough error-corrected machine, break the public-key cryptography that protects much internet traffic. Such a machine does not exist, but governments are already preparing, because data encrypted today could be stored and decrypted later. Our article on post-quantum encryption timelines covers the standards and deadlines that followed.
Reading the headlines
A few questions cut through hype. Are the qubits physical or logical? What is the error rate, and how long do the qubits stay stable? What was actually computed, and was it compared fairly with the best classical method? Is it a research result or a product? A big qubit count alone says little about how capable a machine is.
