What is a Quantum Computer?

qc/basics qc/hardware

A quantum computer is a device that stores and transforms information in qubits — quantum-mechanical states — instead of classical bits, exploiting superposition and entanglement to represent and manipulate information in ways classical bits fundamentally can’t. Key insight: it is not a faster classical computer. A classical bit is always definitely 0 or 1; a qubit’s advantage comes from interference between possibilities, not from “trying every answer at once for free” — you still only ever read out plain classical bits at the end (see Measurement and Collapse).

What’s physically inside one

IBM’s devices (and the ones used throughout this vault) use superconducting transmon qubits — tiny circuits etched from superconducting metal on a chip, each behaving like an artificial atom with quantized energy levels that stand in for $|0⟩$ and $|1⟩$. Other physical implementations exist — trapped ions (individual atoms held by electromagnetic fields, controlled with lasers), photonic qubits (single photons) — but superconducting is what Heron/Nighthawk-class IBM hardware uses.

What’s actually inside a transmon: a Josephson junction — a thin insulating gap between two superconductors — acting as a nonlinear inductor in an otherwise ordinary LC oscillator circuit. The junction’s nonlinearity is essential: a plain (linear) LC oscillator has evenly-spaced energy levels, so no pair of levels could ever be addressed individually — every drive pulse would excite all levels at once. The Josephson junction’s anharmonicity unevenly spaces the levels, which is precisely what lets $|0⟩$ and $|1⟩$ be isolated and driven independently without leaking into higher levels. The macroscopic quantum variable being manipulated is the superconducting phase $\varphi$ across the junction.

Platform	Qubit encoding	Dominant noise source	Strengths
Superconducting (IBM)	Josephson-junction anharmonic oscillator	Decoherence (see Physical Decoherence Mechanisms)	Fast gates, scalable fabrication
Trapped ion	Hyperfine atomic states	Motional heating, laser noise	Long coherence, all-to-all connectivity
Neutral atom	Atomic states, optical tweezers	Atom loss, laser noise	Flexible qubit layout
Photonic	Photon states	Photon loss	Room-temperature operation, natural for networking

Why it needs to be so cold and isolated

Superconducting qubits only behave quantum-mechanically (hold superposition without immediately decaying) when isolated from thermal noise and stray electromagnetic fields. IBM’s chips run inside a dilution refrigerator at around 15 millikelvin — colder than deep space. Any stray heat or noise leaking in causes the qubit to lose its quantum state (decoherence), which is the root cause behind $T_1$/$T_2$ limits and gate errors (see Backend Properties).

The architecture, top to bottom

• Room-temperature classical electronics — generate and read microwave control pulses; a classical computer still orchestrates the whole experiment.
• Cryostat, staged down to ~15 mK — the QPU sits at the coldest stage; wiring runs down through intermediate temperature stages to minimize heat leaking in.
• The QPU chip itself — the physical qubits, laid out in a fixed grid dictated by chip fabrication.

The QPU is a specialized co-processor a classical computer calls out to, not a standalone replacement for one.

Why connectivity is limited

Physical qubits sit at fixed positions on the chip, wired to interact directly with only their physical neighbors. This fixed wiring is the coupling map — a hardware fact, not a software choice. See Coupling Map and Topology and Heavy-Hex Topology for how that graph looks in practice and what it costs when your circuit needs qubits that aren’t neighbors.

A brief history

1980–82: Paul Benioff shows computation can be built from purely quantum-mechanical laws (a quantum Turing machine); Richard Feynman and Yuri Manin independently propose using quantum systems to simulate nature, since classical computers scale so badly at it (see Hamiltonian Simulation — Why It’s Hard). 1985: David Deutsch formalizes universal quantum computation and proves a quantum machine can simulate any other physical system — see Deutsch’s Algorithm for the first concrete algorithm this produced. 1994: Shor’s algorithm demonstrates an exponential speedup for factoring, proving RSA breakable in principle — named (not derived) in The Quantum Algorithm Zoo. 1996: Grover’s algorithm gives a quadratic speedup for unstructured search, derived in full there.

Related

• What is a Qubit
• Coupling Map and Topology — the physical connectivity constraint described here, in full detail
• Backend Properties — the measurable numbers (error rates, $T_1$/$T_2$) that this physical reality produces
• Superposition
• Deutsch’s Algorithm, The Quantum Algorithm Zoo
• Physical Decoherence Mechanisms

Self-Check

• Could you explain to a friend why a quantum computer isn’t “just a faster classical computer”?
• Why does a superconducting QPU need to be cooled to ~15 millikelvin?
• Why does the coupling map exist at all — what physical fact forces it to?
• Who first proposed using quantum systems to simulate nature, and why?
• Why does a Josephson junction’s nonlinearity matter for being able to address $|0⟩$ and $|1⟩$ individually?