Free · No signup · Open methodology

What will your quantum algorithm actually cost to run?

Estimate logical qubits, surface-code distance, physical qubits, magic-state factories and wall-clock runtime for fault-tolerant quantum algorithms - in your browser, using published scaling models from Fowler, Gidney-Ekerå, Litinski and others.

Algorithm preset

Logical qubits

Toffoli gate count

T-depth (serial layers; parallel layouts reduce this)

Physical error rate

Code cycle time

Error budget

Results update live. Every number is an order-of-magnitude planning estimate - read the methodology before quoting it.

Physical qubits (total)

Estimated runtime

Code distance

Surface-code rounds of redundancy

Data-block physical qubits

2 × logical × d²

T-gate count

4 T per Toffoli (standard decomposition)

Physical per logical

Error-correction overhead factor

Where the qubits go

Highlighted share: magic-state factories. Distillation factories often dominate the machine - they make the T states every non-Clifford gate consumes.

Why resource estimation matters

Quantum computing has moved from lab demos to roadmaps. More than 300 companies are piloting quantum applications, and the enterprise market is projected to grow from $4.8B (2025) to $18.5B (2030). The question every roadmap hits is the same: how many physical qubits, and how long, does a useful algorithm actually need? Resource estimation answers it before the hardware exists - it is how Microsoft sizes its Azure Quantum targets, how Google framed its RSA-2048 studies, and how an entire research workshop series (QRE) stays busy.

What this tool does

Presets from the literature - Shor factoring of RSA-1024/2048/4096 (Gidney & Ekerå 2019), Grover key search on AES-128, and FeMoCo nitrogenase chemistry (Reiher et al. 2017), plus a fully custom mode.
Error correction, honestly - surface-code distance is solved from your physical error rate and error budget using the standard Fowler et al. logical-error scaling, not hand-waved.
Factories included - magic-state distillation is usually the biggest part of the machine. We show a lean estimate (Litinski-style 15-to-1) and a conservative factory-dominated one.
Ranges, not fake precision - every result is a bracket with its assumptions exposed. Real studies differ by 10-100x on the same algorithm; that spread is the information.

Frequently asked questions

Is this a replacement for Microsoft's Azure Quantum Resource Estimator?: No. Microsoft's tool compiles your actual program through Q#/Qiskit and models specific hardware profiles. This tool is the fast, web-scale version: literature presets and standard scaling formulas for planning, teaching and sanity-checking. Use ours to get oriented; use theirs when you have code.
Why do my numbers differ from paper X?: Published estimates differ by orders of magnitude because they assume different error rates, cycle times, distillation schemes and gate synthesis tricks. Gidney-Ekerå needed 20 million qubits for RSA-2048 in 2019; Gidney's 2025 update cut that below 1 million. Change the physical error rate here from 0.1% to 0.01% and watch the same effect.
Can I estimate my own algorithm?: Yes - choose Custom and enter your logical qubit count, Toffoli count and T-depth. Getting those three numbers from your algorithm is the hard part; typical sources are your compiler's resource report or the paper the algorithm came from.
What does "above threshold" mean?: If the physical error rate is worse than the surface code's ~1% threshold, adding redundancy makes things worse, not better. No code distance saves you - the hardware has to improve first.