July 27, 2026 · Tags: quantum computing, error correction, physics
You hear about quantum computing constantly. Someone says it will break all encryption. Someone else says it's 10 years away. Both are right. The gap between those two truths is where the real story lives.
A quantum computer is not a faster classical computer. It works on different physics. A classical bit is strictly 0 or 1. A qubit can exist in both states at once. Two qubits can represent all four possible combinations of 0s and 1s simultaneously. With 50 qubits, the state space holds about a quadrillion possibilities. That exponential scaling is the whole reason anyone is building these machines.
But the popular idea that quantum computers "try all answers at once" is wrong. They manipulate probability amplitudes and use interference to cancel wrong answers while amplifying the right one. When you measure the qubits, you get one answer. Good algorithms make sure it's the right one with high probability.
Actual computation has three steps. Initialize all qubits to a known state. Apply a sequence of quantum gates -- Hadamard for superposition, CNOT for entanglement, phase rotations to steer amplitudes. This sequence is the quantum circuit. Then measure the qubits and collect classical bits. Run the circuit thousands of times and aggregate the results.
Two algorithms get most of the attention. Shor's algorithm factors large integers in polynomial time, which breaks RSA encryption. Grover's algorithm gives a quadratic speedup for search. The most practical near-term application is probably quantum simulation -- using one quantum system to model another, which matters for drug discovery and materials science. Feynmanpedia has a good technical breakdown of how this fits together.
The Four Things Holding It Back #
Physical qubits are fragile. A stray field, a temperature fluctuation, or even a cosmic ray can destroy superposition. That is why superconducting qubits operate at 10 millikelvin, colder than deep space. Current gate error rates are around 10^{-3} to 10^{-4}. Useful computation needs 10^{-12} or better. That is a gap of eight to nine orders of magnitude.
Quantum error correction has enormous overhead. The solution is to encode one logical qubit across many physical qubits using error-correcting codes. Google demonstrated surface code operation below the error threshold in 2025. IBM's [[144,12,12]] gross code needs 288 physical qubits for 12 logical qubits, about 10x more efficient than the surface code but still huge. Current estimates put the physical-to-logical ratio at roughly 1000:1. A useful algorithm needing 1,000 logical qubits would require a million physical qubits.
The classical infrastructure can't keep up. As the machine grows, the decoder processing error syndromes must handle data rates up to 10 terabytes per second. Google's AlphaQubit2 and IBM's Relay-BP decoder are making progress, but distributed decoding across millions of qubits is an unsolved architecture problem. IBM's 2029 roadmap targets a 200-logical-qubit system called Starling but acknowledges the classical control path as a first-order constraint.
Drift means you cannot leave it alone. Qubit calibration drifts over hours. Current practice stops the computation to recalibrate, which is fine for a 20-minute experiment but useless for multi-day algorithm runs. Google published results in Nature (July 2026) showing reinforcement learning can adjust control parameters on the fly using error syndromes as a learning signal. It works on Willow hardware and improves stability 3.5-fold. But this is early research.
Where We Actually Stand #
The field is leaving the NISQ era (noisy intermediate-scale quantum, coined by John Preskill in 2018) and entering "early fault tolerance" -- systems with small numbers of error-corrected logical qubits but tight constraints on latency and connectivity.
Industry roadmaps cluster around the late 2020s for the first genuinely useful machines. IBM's Starling (2029), IonQ's 80,000-logical-qubit target (2030), Quantinuum's millions-of-gates roadmap (2029). Most independent experts I follow put the timeline at 5 to 15 years before fully error-corrected machines with thousands of logical qubits exist.
There is also a question the field does not talk about enough: what will we actually do with these machines? Shor's algorithm breaks RSA. Quantum simulation is useful for chemistry. Beyond that, the list of real-world problems where quantum advantage holds under practical constraints is short. NIST's quantum computing overview frames this better than most industry press releases.
Why This Matters #
Quantum computers will not replace your laptop. They will be specialized accelerators the way GPUs sit alongside CPUs. But the problems they solve -- breaking encryption, designing new drugs, simulating materials -- are big enough to reshape industries.
The cryptographic transition is the most immediate risk. Adversaries can store encrypted data today and decrypt it when a capable quantum machine exists. "Harvest now, decrypt later" is a real threat, and the migration to post-quantum cryptography is slow. NIST published four post-quantum standards in 2024, but the internet's encryption infrastructure does not upgrade quickly.
The machines are coming. Just not as fast as the hype says, and not for the reasons the press releases give you.