How Quantum Computing Works

In 2019, Google announced that its 53-qubit Sycamore processor completed a calculation in 200 seconds that would take the world’s best supercomputer roughly 10,000 years. IBM, irritated, said the number was wrong. A well-tuned classical approach could do it in about 2.5 days. Both claims were technically defensible. The fight revealed something important: quantum computing is real, it works, and we are still arguing about what it actually does.

The short answer

A quantum computer exploits quantum mechanical phenomena: superposition (a qubit can represent 0, 1, or both simultaneously), entanglement (qubits can be linked so measuring one instantly affects the other), and interference (bad answers cancel out, good answers amplify), to explore many solutions at once for specific types of problems. It is not a faster classical computer. It is a fundamentally different kind of machine that is useful for a narrow but important set of tasks.

The full picture

Why bits hit a ceiling

Your laptop runs on bits, tiny switches, that are either off (0) or on (1). Everything a computer does, from loading a webpage to training an AI model, reduces to manipulating billions of these switches.

The problem: for certain problems, the number of states you need to explore grows exponentially with the size of the input.

Factoring a 2,048-bit number (the kind used in RSA encryption) requires trying roughly 2^1024 possible solutions. There isn’t enough time in the age of the universe for a classical computer to brute-force that, even at trillions of operations per second.

Quantum computers don’t speed up the search. They change the search entirely.

Qubits: not magic, but close

A classical bit must be 0 or 1 at any given moment. A qubit (quantum bit) can exist in a superposition of both states simultaneously, described as a combination: some probability of being 0, some probability of being 1, when measured.

The crucial word is when measured. Before measurement, a qubit isn’t secretly one or the other: it genuinely exists in both states at once: not a metaphor or a shortcut but how quantum systems actually behave, a fact that baffled physicists for decades and is still philosophically strange.

Superposition gives you leverage. With 2 qubits, you can simultaneously represent 4 states (00, 01, 10, 11). With 10 qubits, 1,024 states. With 300 qubits, more states than there are atoms in the observable universe. The processing space scales exponentially.

But here’s the catch: when you measure a qubit, superposition collapses. You get a single answer: 0 or 1. A quantum computer can’t just “read out” all the answers at once. The art of quantum computing is in setting up the computation so that the answer you want is overwhelmingly likely to survive measurement.

Entanglement: spooky action at a distance

Einstein called it “spooky action at a distance.” Physicists call it entanglement.

When two qubits are entangled, their states become correlated in a way that has no classical equivalent. Measuring one qubit instantly determines something about the other, regardless of the physical distance between them.

This isn’t communication or information transfer. You can’t use entanglement to send data faster than light, but it does allow quantum computers to link qubits together in ways that let operations on one qubit instantly affect another. In a 53-qubit processor like Google’s Sycamore, entanglement means the qubits don’t act as 53 independent switches but as a single, deeply interconnected quantum system.

Entanglement is part of why quantum computers can explore solution spaces exponentially larger than classical machines. The qubits aren’t computing independently; they’re computing together.

Interference: how you get useful answers

Superposition lets a quantum computer explore many possible answers simultaneously. Entanglement links qubits into a coherent system. But neither of these explains how you extract a correct answer.

The answer is interference.

Quantum states can interfere with each other the way waves interfere. Paths leading to correct answers can be amplified (constructive interference). Paths leading to wrong answers can be cancelled out (destructive interference).

A quantum algorithm is, at its core, a carefully designed interference pattern. The algorithm sets up a quantum state, applies a sequence of operations (called quantum gates), and steers the interference so that when you finally measure the qubits, the measurement is overwhelmingly likely to return the right answer.

This is why quantum computing isn’t just “trying all answers at once.” A naive attempt to do that would collapse to a random answer on measurement. Quantum algorithms are precise interference engineering.

The most famous example: Grover’s algorithm searches an unsorted database of N items in roughly √N steps instead of N/2 steps. On a database of a million items, that’s the difference between 500,000 operations and 1,000 (a 500x speedup). More powerful: Shor’s algorithm factors large numbers in polynomial time rather than the exponential time required classically, which is why quantum computing is considered an existential threat to current encryption.

The hardware problem: fragile by nature

Quantum states are delicate. Superposition and entanglement require qubits to remain isolated from environmental disturbances: heat, vibration, stray electromagnetic fields, even a single photon hitting the system. Any interaction with the outside world causes decoherence: the quantum state collapses, superposition vanishes, and you’re left with noise.

This is why quantum computers currently operate at temperatures near absolute zero (around 15 millikelvin, colder than outer space, inside elaborate refrigeration systems about the size of a chandelier).

Maintaining coherence for long enough to run a useful calculation is the central engineering challenge of quantum computing. Current machines operate for microseconds to milliseconds before decoherence destroys the computation.

The answer is error correction, but quantum error correction is vastly harder than classical error correction. Classical computers can simply copy bits and take a majority vote. Copying quantum states is physically impossible (the no-cloning theorem). Quantum error correction instead encodes one logical qubit across many physical qubits, using entanglement to detect and fix errors without disturbing the computation.

Current estimates suggest a useful fault-tolerant quantum computer would need hundreds of logical qubits, each encoded across hundreds or thousands of physical qubits, meaning millions of physical qubits working together. The largest systems today have hundreds of physical qubits with limited error correction. The gap is significant.

Different types of quantum computers

Not all quantum computers use the same technology to build qubits. The main approaches:

Superconducting qubits: used by Google, IBM, and Rigetti. Tiny circuits of superconducting metal cooled to near absolute zero. Fast gate operations but fragile. Current machines: IBM’s Heron (133 qubits), Google’s Willow (105 qubits, announced December 2024 with significantly improved error rates).

Trapped ions: used by IonQ, Quantinuum (formerly Honeywell). Individual ions suspended in electromagnetic fields. Slower but more accurate. IonQ’s Forte has 36 algorithmic qubits (a quality-adjusted metric).

Photonic qubits: used by PsiQuantum, Xanadu. Encode information in photons. Operate at room temperature, but hard to entangle reliably. Requires massive numbers of components.

Topological qubits: Microsoft’s long-bet approach. Theoretically much more stable, but Microsoft’s first physical demonstration of the underlying particle (the Majorana fermion) only came in 2025. Years from practical deployment.

Each approach makes different engineering trade-offs between qubit count, error rate, speed, and operational temperature.

What quantum computers are actually good for

The most practical near-term applications:

Quantum chemistry simulation: Simulating the quantum behavior of molecules is intractable for classical computers (nature is quantum, and you can’t efficiently simulate quantum systems classically). A quantum computer could model drug interactions at the molecular level, discover new materials for batteries or solar cells, or design more efficient fertilizer production processes. This could cut years off pharmaceutical development timelines.

Optimization problems: Finding the optimal solution among exponentially many options: supply chain routing, financial portfolio optimization, traffic flow, protein folding. Quantum approaches like the Quantum Approximate Optimization Algorithm (QAOA) may offer advantages for some of these problems, though proof of consistent real-world speedups remains elusive.

Cryptography: Specifically, breaking RSA and elliptic curve encryption using Shor’s algorithm. Also enabling quantum key distribution (QKD), which is theoretically unbreakable because eavesdropping disturbs the quantum state and reveals itself.

Machine learning: Quantum-enhanced machine learning remains speculative. Some quantum linear algebra operations could theoretically accelerate certain ML tasks, but the data-loading problem (getting data into a quantum computer efficiently) may offset any gains.

The encryption threat and the response

RSA encryption secures HTTPS websites, banking systems, email, and most of the internet. It works because multiplying two large prime numbers together is easy, but factoring the result is computationally hard.

Shor’s algorithm makes factoring efficient on a quantum computer. A large enough quantum computer could break RSA in hours.

“Large enough” is the key phrase. Breaking 2,048-bit RSA would require roughly 4,000 logical qubits, each protected by high-fidelity error correction, implying millions of physical qubits. Current machines are thousands of times too small.

But “harvest now, decrypt later” attacks are already underway. Nation-state adversaries are believed to be collecting encrypted internet traffic today, storing it, and waiting for quantum computers powerful enough to decrypt it. Any information that needs to remain secret for 10-15 years is potentially at risk.

In response, NIST (the U.S. National Institute of Standards and Technology) finalized its first post-quantum cryptography standards in August 2024: ML-KEM (CRYSTALS-Kyber), ML-DSA (CRYSTALS-Dilithium), and SLH-DSA (SPHINCS+). These algorithms are believed to resist both classical and quantum attacks. Migration to post-quantum cryptography is already underway in governments and large enterprises.

Where things stand in 2026

The field is advancing faster than most experts predicted five years ago. Google’s Willow chip (December 2024) demonstrated exponentially better error correction as qubit count increased, the first time scaling actually reduced error rates rather than increasing them, a key milestone. IBM is on a roadmap to utility-scale quantum computing by the end of the decade.

But “faster progress than expected” still means the gap to fault-tolerant, cryptography-breaking quantum computers is measured in years to decades, not months.

The near-term reality: quantum computers are research tools, useful for probing quantum chemistry and exploring algorithms. They are not yet better than classical computers at any commercially important problem. The medium-term reality: the first quantum advantage in chemistry or optimization is likely in the early-to-mid 2030s. The long-term: if the engineering problems get solved, the consequences for cryptography and drug discovery are profound.

This is one of the few technologies where the hype is, if anything, a slight understatement about the long run, and a massive overstatement about the short run.

Common misconceptions

“A quantum computer is just a faster classical computer.” It is not faster at everything. A quantum computer running a word processor or streaming video would be slower than your laptop. The advantage is specific: certain mathematical structures that are intractable for classical machines become solvable. Speed is the wrong frame. Capability on specific problem types is the right one.

“Quantum computers will break all encryption soon.” Not soon. Breaking RSA encryption with Shor’s algorithm requires millions of error-corrected logical qubits. Today’s best machines have hundreds of noisy physical qubits. The threat is real and governments are preparing for it, but “soon” means decades away, not years. The more immediate concern is “harvest now, decrypt later”: adversaries collecting encrypted data today to crack once capable machines exist.

“Quantum computers work by trying all answers simultaneously.” This is the most common oversimplification and it is wrong in an important way. Quantum computers do not try every possibility at once and pick the winner. They manipulate probability amplitudes so that wrong answers cancel out and correct answers reinforce. Without that interference step, measuring a qubit gives you a random result, not the right one. The computation is in the choreography of the wave, not in parallelism.

Key terms

• Qubit: The quantum equivalent of a classical bit, capable of superposition (existing as 0, 1, or both)
• Superposition: A quantum state where a system exists in multiple states simultaneously until measured
• Entanglement: A correlation between qubits such that the state of one instantly relates to the state of another
• Interference: The quantum phenomenon used to amplify correct answers and cancel wrong ones
• Decoherence: The collapse of a quantum state due to interaction with the environment
• Shor’s algorithm: A quantum algorithm that factors large numbers exponentially faster than classical methods
• Grover’s algorithm: A quantum algorithm that searches unsorted databases with a quadratic speedup
• Logical qubit: A fault-tolerant qubit encoded across many physical qubits for error correction
• Post-quantum cryptography: Classical encryption algorithms designed to resist quantum attacks