Cost Models & Access Patterns

Overview

This page answers the question: “Even if a device exists, what does it cost to use it in practice?”

Hardware performance isn’t only physics. Access patterns—how you submit jobs, how long you wait, what you pay for, and how reproducible results are—shape what you can realistically do.

Intuition

There are at least three kinds of “cost” in quantum experiments:

• Money cost: what you pay to run workloads.
• Time cost: waiting for queues, calibrations, retries, and debugging cycles.
• Iteration cost: how fast you can learn, change a circuit, and run again.

A device can be “cheap per run” but still expensive in time if the queue is long. Or it can be fast to access but expensive to run at scale.

This connects to Noise & Errors because noisy workloads often require repetition:

• you run many shots to estimate expectation values
• you repeat experiments to check stability over time
• you may use mitigation workflows that increase run count

So access and pricing models shape what mitigation and validation are feasible.

What This Metric Captures

Cost and access descriptions often capture:

• Cloud access vs on-prem access: whether you submit jobs remotely or control the environment locally.
• Shot-based usage: charging or quotas based on repeated measurements, which is natural for probabilistic outputs.
• Queueing and scheduling: whether jobs run immediately or wait behind other users/workloads.
• Availability windows: when the system is online, stable, or undergoing calibration/maintenance.

These are not purely economic details; they are constraints on experiment design.

What This Metric Misses

A simple “price per shot” (or any single cost figure) misses important realities:

• End-to-end workflow time: compilation, validation runs, debugging, and reruns are often the real bottleneck.
• Variance in results: if outcomes drift, you may spend more on repeated verification.
• Fair comparison across workloads: some workloads need many short jobs; others need fewer but longer or more constrained jobs.
• Hidden constraints: limits on job size, concurrency, calibration states, or allowed operations can matter more than price.

In practice, cost is about the total effort to get a reliable scientific or engineering answer—not just the amount billed.

Turtle Tip

Common Pitfalls

Quick Check

What’s Next

You can now interpret key “spec sheet” dimensions: qubit count, connectivity, gate quality, and access cost. Next we’ll step up a level and discuss provider ecosystems—how hardware is packaged, what abstractions matter, and how to stay provider-agnostic while still being practical.