Qubit Count

Qubit count refers to the total number of qubits available on a quantum processor. It is the most frequently reported specification in quantum computing announcements, but it is also the most misleading when used in isolation. A processor with 1,000 noisy qubits may be less capable than one with 100 high-quality qubits, because the depth of circuits that can run successfully depends on error rates, connectivity, and coherence times, not just the number of qubits.

Why raw qubit count is misleading

Consider a simple analogy: the number of transistors in a classical processor matters, but a chip with billions of unreliable transistors would be useless. Similarly, a quantum processor’s utility depends on what computations its qubits can successfully perform, not merely how many qubits exist.

Several factors make raw qubit count a poor predictor of capability:

• Gate errors: If two-qubit gate error rates are $1\%$, a circuit with 100 sequential two-qubit gates has an expected error rate of approximately $1 - (0.99)^{100} \approx 63\%$. More qubits enable wider circuits, but errors accumulate with depth.
• Connectivity: If qubits are arranged in a line (nearest-neighbor connectivity), moving information between distant qubits requires many SWAP gates, each adding noise. A fully connected processor can run the same algorithm with far fewer gates.
• Coherence time: Qubits that decohere in $50\ \mu s$ can execute fewer gates than qubits that remain coherent for $1\ ms$, regardless of how many qubits exist on the chip.
• Crosstalk: On densely packed processors, operating one qubit can disturb its neighbors. More qubits in a smaller area can increase crosstalk, degrading overall performance.

Better metrics

The quantum computing community has developed several metrics that capture more than just qubit count:

• Quantum Volume (QV): Measures the largest square random circuit (width $n$, depth $n$) a processor can run with success probability above $2/3$. A QV of $2^n$ means the processor can reliably handle $n$-qubit, $n$-depth circuits.

• Circuit Layer Operations Per Second (CLOPS): Measures how quickly a processor can execute parameterized circuits, capturing both quantum gate speed and classical control overhead.

• Algorithmic qubits: A metric proposed by some vendors to estimate how many “useful” qubits a processor provides for specific algorithms after accounting for error rates and connectivity.

• Randomized benchmarking: Measures average gate fidelity by running random Clifford sequences, providing a hardware-agnostic characterization of gate quality.

No single metric captures all relevant dimensions, which is why benchmarking quantum computers remains an active area of research.

Current state of qubit counts

As of early 2026:

Vendor	Platform	Qubit count	Notable quality metric
IBM	Superconducting (Heron)	156	QV 256+
IBM	Superconducting (Condor)	1,121	Technology demonstration
Google	Superconducting (Willow)	105	Below-threshold surface code
Quantinuum	Trapped ion (H2)	56	QV 65,536+
QuEra	Neutral atoms	256+	High two-qubit gate fidelity
IonQ	Trapped ion (Forte)	36	High algorithmic qubit count

The contrast between IBM’s 1,121-qubit Condor and Quantinuum’s 56-qubit H2 is instructive. Quantinuum’s system, despite having 20 times fewer qubits, achieves a quantum volume many orders of magnitude higher because its trapped-ion qubits have better fidelity and all-to-all connectivity.

Qubit count and fault tolerance

For fault-tolerant quantum computing, the relevant question shifts from “how many physical qubits?” to “how many logical qubits at what error rate?” Running Shor’s algorithm to break RSA-2048 is estimated to require roughly 4,000 logical qubits, each encoded in approximately 1,000 physical qubits (using the surface code at distance 25 with physical error rates around $0.1\%$). This puts the total physical qubit requirement in the millions, far beyond current systems.

Why it matters for learners

Qubit count is the quantum computing metric most likely to appear in news headlines, and understanding its limitations is essential for evaluating claims and progress. When a company announces a new processor, the questions to ask are: What are the two-qubit gate fidelities? What is the quantum volume or equivalent benchmark? What connectivity does the chip provide? These details, not the raw qubit count, determine whether the processor represents a meaningful advance.

See also

• Quantum Volume
• NISQ
• Physical Qubit
• Randomized Benchmarking