What does it really take to run a quantum computation?

Qubits are extremely fragile. Heat, vibration, stray magnetic fields — almost anything disturbs them. Even the best machines make roughly one error per thousand operations. Useful algorithms need billions of operations. Without correction, the result would be pure noise long before the computation finishes.

The trick is redundancy: many fragile physical qubits are woven together so errors can be detected and fixed on the fly, creating one reliable "logical" qubit. How many you need depends on how error-prone the hardware is. Worse qubits — more redundancy — bigger machines. That is the single most important lever in the entire field — and why hardware quality (fidelity) matters more than raw qubit count.

Small demonstrations with a handful of qubits genuinely run today — that is real. But there is a gulf between a 4-qubit chemistry demo and the millions of physical qubits needed to break encryption or design an industrial catalyst. Both statements are true at once: the technology works, and the commercially decisive applications are still years of engineering away. Serious roadmaps and honest resource estimates — like this one — are how you tell hype from progress.