threshold theorem

Real qubits are noisy. Every gate you apply, every moment a qubit just sits there waiting, adds a little error. So here is the obvious worry: if you build a bigger computer with more qubits and longer programs, don't the errors just pile up until the answer is garbage? The threshold theorem is the mathematical result that says no, not necessarily. It proves that if the error rate of your basic physical operations is below a certain cutoff value, called the threshold, then you can use quantum error correction to push the effective error rate of your computation as low as you like. Think of it like noise-canceling on a phone call: below a certain level of background noise the system can keep cleaning up faster than the noise creeps in, so a long call stays clear.

The catch is in the word below. Error correction spends extra physical qubits and extra operations to protect a small amount of real information (one protected, or logical, qubit). If your hardware is already cleaner than the threshold, adding more layers of protection makes things steadily better, and the cost grows only modestly as you demand higher reliability. If your hardware is dirtier than the threshold, adding protection makes things worse, because the correcting machinery introduces more errors than it removes. For the surface code, the most studied scheme, the threshold sits around 1% error per operation, which today's best hardware is only beginning to reach.

So the threshold theorem is the reason most researchers believe large, reliable quantum computers are possible in principle rather than ruled out by noise. But it is a statement about what is achievable, not a free lunch. Staying below threshold is hard, and the price of protecting even one logical qubit is often hundreds or thousands of physical qubits. That overhead is exactly why we are still in the NISQ era today, with no large-scale fault-tolerant machine yet built.