The State of Quantum Computing
Companies announce record-breaking quantum machines. Some promise breakthroughs in drug discovery, cryptography, and materials science. But how close are we, really, and what does “working quantum computer” even mean? This overview uses real hardware data and fundamentals to answer those questions.
Why would a quantum computer be faster?
Start with what “faster” means. Every problem, sorting a list, finding a route, breaking a code, is solved by an algorithm: a step-by-step recipe. If a quantum algorithm solves a problem in 1,000 steps where the best classical algorithm needs a billion, the quantum computer wins. The advantage is in the algorithm, not the clock speed.
Quantum computers can execute certain algorithms with dramatically fewer steps because they exploit three properties of physics that classical computers cannot access:
Qubits are extraordinarily fragile
Each step in a quantum algorithm is called a gate, the quantum equivalent of a classical logic operation, like an AND or NOT gate. The problem is that every gate carries a small probability of going wrong. Heat, vibration, stray electromagnetic fields, any disturbance can nudge a qubit away from its intended state, corrupting the result.
These errors accumulate. Imagine passing a secret message along a chain of people, where each person has a small chance of mishearing a word. Pass the message through ten people and you’re probably fine. Pass it through ten thousand and the message is unrecognisable.
You might ask: why not just build qubits that don’t make errors? Think of friction. No mechanical system is perfectly smooth, and eliminating friction entirely would change the system fundamentally. Qubits face the same constraint: any qubit sensitive enough to hold quantum information is equally sensitive to every stray disturbance from its environment. There is no known way to shield one completely without also stopping it from doing useful work.
Where we are today
3 hardware technologies are currently leading the race. Each takes a different physical approach to building qubits, with different trade-offs between qubit count and error rate.
Data represents the current best-demonstrated performance for each hardware type. Values update as new records are set.
The best machines today achieve error rates around 3×10⁻⁴, roughly 0.03% per gate. That sounds small, but errors compound: across 10,000 gates, the probability of at least one mistake is already above 95%. Meaningful quantum algorithms need millions or billions of gates, and at current error rates, the result would be pure noise.
The good news: error rates have been falling steadily.
What would it actually take to do something useful?
One of the most-cited quantum applications is Factoring RSA-2048, breaking the RSA encryption standard still widely used in digital certificates and key exchange. A quantum algorithm called Shor’s algorithm can find the prime factors of a large number exponentially faster than any known classical method. In practice: the best classical computers would need longer than the age of the universe to factor a 2048-bit RSA number. A sufficiently powerful quantum computer could do it in hours.
An error-corrected qubit, also called a logical qubit, is built from many physical qubits working together to actively detect and fix errors. Resource estimates for Shor’s algorithm are given in logical qubits, since raw physical qubits accumulate too many errors to sustain a billion-gate computation.
Resource estimates from published research.
Today’s largest quantum processor has 260 physical qubits, nowhere near 1,399 logical ones. And even with enough qubits, running 6.5×10⁹ Toffoli Gates gates at today’s error rates would bury the result in noise. The solution is error correction.
Error correction: the fix with a cost
Physicists have known since the 1990s that quantum errors can be corrected without disturbing the underlying quantum information, a surprising result. Measuring a qubit is like opening Schrödinger’s box: the act of looking forces it to pick a definite value, collapsing the superposition that made it useful. Error correction sidesteps this by measuring the relationships between qubits indirectly, not the qubits themselves. The trick is redundancy: spread one logical qubit of information across many physical qubits, and errors in a few can be detected and corrected without ever disturbing the quantum state directly.
The surface code is today’s most widely used approach, reliable and well-understood, the established workhorse of the field. It requires roughly a thousand or more physical qubits to protect a single logical qubit. Newer schemes called qLDPC codes, quantum adaptations of a classical family of efficient error-correcting codes, are being actively developed and promise to reduce that overhead substantially. The leading candidate, the [[144, 12, 12]] Bivariate Bicycle code, uses only 31.5 physical qubits per logical qubit.
Even at 31.5 physical qubits per logical qubit, Factoring RSA-2048 would require roughly 44,226 physical qubits. Today’s largest machines have 260.
Qubit counts have been growing steadily, though the pace varies by hardware type:
So when does quantum computing arrive?
Quantum computers already exist. What hasn’t arrived yet is a clear, demonstrated advantage over the best classical computers for a problem that actually matters, what researchers call quantum advantage.
The gap is closing from both directions. Hardware error rates keep falling and qubit counts keep rising, as the charts above show. On the software side, better algorithms and more efficient error correction schemes are steadily reducing what a useful computation requires. Both frontiers are active, and both are moving. It is a target that keeps getting closer.
That’s what this site tracks. Explore the numbers yourself: Guided exploration for a walkthrough, or Expert to run your own scenarios.