The Open Problems

We have the algorithms and error correction in principle — the wall is scaling up qubits while keeping errors low, a brutal but not impossible engineering challenge.

Quantum Computing · Lesson 19 · 10 min read

We know the algorithms — Shor, Grover, simulation. We know error correction works in principle. We even have working small machines. So why don’t we have a useful quantum computer yet? It’s not one dramatic obstacle but a stack of brutal engineering problems that all get harder the more qubits you add. Hold the question: what actually stands between today’s noisy hundred-qubit machines and the million-qubit, fault-tolerant ones we’d need — and is the wall fundamental, or just hard?

The core wall: scale vs. error

Error correction only helps when physical error rates are below threshold, and a useful machine needs on the order of millions of physical qubits because of the overhead (L15). But every qubit you add brings more noise, more ways to fail, and more control complexity. So scaling up and keeping error rates down pull against each other — it’s the L13 isolation-vs-control tension, multiplied a millionfold. That conflict is the heart of the problem.

The specific hard problems

Concretely: (a) gate error rates must drop further and stay low at scale; (b) crosstalk — packed qubits disturb their neighbors, so more qubits can mean more errors; (c) the wiring/control problem — each qubit needs precise control lines and (often) extreme cooling, so millions of them is a staggering plumbing-and-electronics challenge; (d) no hardware platform (superconducting, trapped ions, photonics, and others) has clearly won yet. Each of these is an active research frontier, not a solved detail.

Hard, but not forbidden (the honest optimism)

Here’s the crucial distinction: no known law of physics forbids a large fault-tolerant quantum computer — unlike, say, perpetual motion, which the laws of thermodynamics rule out. The obstacles are engineering (and possibly better error-correcting codes), and every one is seeing steady progress. So the honest stance is “very hard, uncertain timeline, but not blocked by nature.” That is a fundamentally different situation from a true impossibility — and the reason serious people remain optimistic without promising a date.

Why “just add more qubits” doesn’t work:
1. Start: ~100 noisy physical qubits, error rate near threshold. Good for shallow demos.
2. Naive plan: add qubits to build the logical qubits error correction needs.
3. Problem 1: each added qubit adds noise and crosstalk, nudging the average error rate UP — possibly above threshold, where error correction actually makes things WORSE.
4. Problem 2: a useful machine needs roughly millions of physical qubits → millions of control lines, massive cooling, and high fabrication yield — a plumbing-and-electronics nightmare.
5. So progress isn’t “add qubits.” It’s “make qubits good enough AND scalable AND well-connected, all at once.” That triple constraint is the real frontier — hard and advancing, but not forbidden by physics.