Why do quantum computers still make mistakes?

Tapasya Pandiyan, GIQ member

Something people don’t really talk about with quantum computers: the hard part isn’t building more qubits, it’s keeping the ones you already have from messing up.

Qubits are extraordinarily fragile. Heat, vibration, even random electromagnetic noise nearby can knock them out of the state you need them in and this happens in microseconds. So basically every calculation is racing against its own hardware falling apart mid computation.

That’s why we’re still stuck in what people call the nisq era (Noisy Intermediate Scale Quantum). Noisy because errors are still everywhere. Intermediate scale because we have enough qubits now to be interesting but nowhere near enough to actually be reliable.

The fix is error correction but it’s brutally expensive to do. To get one logical qubit , you need dozens or even hundreds of physical qubits all linked together, constantly checking and fixing each other’s mistakes. So a chip with 1,000 physical qubits might only give you a handful of qubits you can actually rely on.

Progress is happening faster than a lot of people expected though Harvard got 48 logical qubits working on a neutral atom setup and some researchers think we might hit around 100 logical qubits by the end of this year, which is a pretty big jump from where things were just a couple years ago

And this isn’t just a lab curiosity. Once error correction actually gets good enough, a lot of the stuff people have been hyping up for years starts to become real. Drug discovery is a big one, since molecules behave quantum mechanically and classical computers can only approximate that behavior. Cryptography too,algorithms like Shor’s could eventually break most of the encryption the internet runs on, which is part of why post quantum cryptography is already being built now, ahead of time. Same idea with designing new materials for things like batteries or solar cells and with optimization problems in logistics or finance that have too many moving parts for classical computers to bruteforce.

None of this works yet in any real world sense. All of it is stuck behind the same wall: getting past nisq. That’s honestly the most interesting part of where the field is right now. The theoretical potential is huge but the actual bottleneck is an engineering problem that’s turning out to be way harder than anyone hoped and most of the real progress happening right now isn’t about adding more qubits, it’s about making the ones we have trustworthy.