Quantum threat timeline · continuously refreshed

How close is quantum hardware to breaking RSA-2048?

A 0–100 score measuring distance to a published resource estimate, computed from vendor specifications and peer-reviewed papers rather than press releases. The headline is the field frontier — the highest score anyone currently holds — not a named winner. The table names every machine and its own score.

Anchor

Gidney 2025

Target

RSA-2048

Logical qubits

4,000

Physical qubits

1,000,000

Runtime

<1 week

Confidence

peer reviewed

Frontier threat score

8.36/100

Confidence band 4.18–8.36

Systems scored

Plus 2 marked N/A — a category difference, not a low score

Composite industry

1.04/100

Mean across every scored system, not the frontier

Below threshold

Systems with a demonstrated below-threshold error-correction result

Why the number is this low, and why that is the honest answerThe score is a multiplicative gate: logical capacity × fidelity gate × error-correction signal. Any component at zero takes the whole score to zero, because breaking RSA requires all of them — there is no “mostly there.” Today no system holds standing, error-corrected logical qubits, so the threat column reads near zero across the board. Readiness is a separate axis and is not distance to breaking RSA — it tracks preconditions assembled. A machine can sit at 80% readiness and zero threat at the same time, which is the honest state of most of the field. We do not publish a projected Q-Day year; a projection is only as good as its trajectory model, and ours is not yet defensible enough to publish.

The anchor is a moving target, and that cuts both waysThis score measures distance to Gidney 2025 — 1,000,000 physical qubits, <1 week. The previous published estimate, Gidney+Ekera 2019, put it at 20,000,000 physical qubits (2019-05). That is roughly a 20× reduction in the requirement in about six years — from algorithmic improvement, not from better hardware. A score against a fixed target would have looked like progress that never happened. If the estimate moves again, every number on this page moves with it, and we will say so rather than quietly restate the scale. Prior estimate ↗ Current anchor ↗

Per-system scoring

Every machine, named, with its own inputs.

Fidelity is deliberately not presented as a ranking — an XEB figure and a median-ECR figure are not the same physical quantity.

System	Modality	Threat	Readiness	Fidelity gate	EC signal	Input confidence
Google WillowReleased 2024	superconducting	8.36	86.7%	100%	100%	high
IBM Heron r2Released 2024	superconducting	0.00	30.5%	58%	15%	low
Quantinuum H2-1Released 2024	trapped_ion	0.00	36.0%	75%	15%	low
IonQ ForteReleased 2023	trapped_ion	0.00	32.8%	69%	15%	low
IQM GarnetReleased 2024	superconducting	0.00	30.9%	66%	15%	medium
Atom Computing next-gen (1,225-site array)Released 2023	neutral_atom	0.00	37.8%	69%	15%	low
Rigetti Ankaa-3Released 2024	superconducting	0.00	26.4%	50%	15%	low
USTC (Zuchongzhi) Zuchongzhi 3.0Released 2025	superconducting	0.00	34.6%	70%	15%	medium

Not scored

Systems outside the gate model.

Analog simulators have no gate-model two-qubit fidelity and no gate-model path to Shor's. Marked N/A rather than scored zero.

System	Modality	Why not scored
QuEra Aquila	neutral_atom	analog_not_gate_model
Pasqal Orion Alpha	neutral_atom	analog_not_gate_model

Sourcing

Where every figure comes from.

Peer-reviewed publications and official vendor technical documents only. Where a press release disagrees with a published figure, the published figure wins.

Full methodology →Anchor paper Dataset source

Corrections

Challenge a number.

A sourced correction improves every subsequent refresh. Concrete beats polite.

Invite scrutiny

See a number you disagree with?

This index is only as good as the scrutiny it survives. If a spec is wrong, a source is weak, a method tag is misapplied, or a system is missing — tell us. Concrete corrections improve every subsequent run. For long-form critique or anything you want a public record of, open a GitHub Issue instead.