Orientation guide
Applications, Evidence, and Strategy
Sooner or later every quantum conversation becomes a decision: build, partner, invest, wait, or walk away. This guide gives you the two tools that make that decision boring in the best way — an evidence ladder for application claims and a decision frame for acting on them — then applies both to chemistry, optimization, machine learning, cryptography, and sensing.
In this chapter 9 sections
Reader question. What evidence is sufficient to build, partner, invest, wait, or avoid for a proposed quantum application?
The decision should follow the buyer's workflow, computational kernel, strongest classical baseline, integration cost, and dated evidence rung; action is justified only when the next proof step is cheaper and more informative than the uncertainty it resolves.
- This guide does not predict market size or treat adjacent quantum sensing as proof of quantum-computing value.
- It does not convert research novelty, small synthetic benchmarks, or vendor access into customer ROI.
Begin at the buyer's blocked workflow
Identify owner, input, output, cost, and decision metric.
Quantum value is not "quantum applied to an industry." It appears when a specific problem has structure that a quantum method exploits better than the best classical workflow, under realistic hardware assumptions, at a cost someone will pay. Every clause of that sentence is a filter, and most claims fail at least one.
Once a claim survives the filters, the strategic question arrives: should we build, partner, invest, monitor, or avoid? Everything in this guide exists to make that answer defensible.
| Level | Evidence | Meaning |
|---|---|---|
| 0 | Narrative only | Interesting, but not decision-grade |
| 1 | Toy demo | Educational, not commercial evidence |
| 2 | Simulation with assumptions | Useful for research; needs resource realism |
| 3 | Hardware experiment | Shows feasibility in a narrow setting |
| 4 | Resource-estimated fault-tolerant workflow | Serious long-term application signal |
| 5 | Beats a strong classical baseline end to end | The strongest technical signal |
| 6 | Repeated customer use with ROI | Commercial validation |
Most computing application claims today sit between levels 1 and 4; real work, honestly done, several rungs short of advantage. Some adjacent quantum technologies, sensing and security migration especially, have nearer-term commercial paths precisely because they do not need a fault-tolerant computer.
Evidence boundary. Quantum application value must be evaluated against end-to-end workflows and classical alternatives. [National Academies of Sciences] [gao-quantum]
Locate the quantum-shaped kernel
Separate a plausible computational mechanism from the surrounding pipeline.
The case for quantum help is structural: molecules and materials are quantum systems, exact classical simulation scales brutally, and better simulation would matter for catalysts, batteries, drugs, and industrial materials.
What to watch in any specific claim:
- the active-space choice,
- the Hamiltonian encoding,
- the target precision,
- logical qubit counts,
- T-gate or non-Clifford cost,
- the classical chemistry baseline,
- integration with HPC and AI workflows.
Strategic view: chemistry and materials are among the strongest long-term application areas, but broadly useful workloads likely require fault tolerance. Long-term strength and near-term revenue are different properties; do not let a pitch blur them.
People care because the money is real: routing, scheduling, portfolio construction, logistics, and design are economically enormous.
Caution is required because the competition is too:
- classical heuristics are extremely strong and improve constantly,
- real problem data and constraints are messy,
- quantum speedups for these problems are often unclear,
- annealing, QAOA, and hybrid methods must each be compared against production-grade baselines.
Strategic view: optimization is commercially attractive and technically dangerous; the area where overclaiming is easiest and baseline discipline matters most.
Evidence boundary. Quantum chemistry is a natural application area but useful results depend on algorithms, resources, and workflow integration. [Yudong Cao et al.] [National Academies of Sciences]
Name the opponent before the experiment
Specify the strongest classical baseline and tuning budget.
People care because machine learning is economically central, quantum feature spaces and kernels are genuinely interesting, and hybrid models are easy to prototype.
Caution is required because:
- data loading can dominate the runtime,
- trainability can fail outright,
- classical models are powerful and improving,
- benchmark leakage and weak baselines are common in the literature.
Strategic view: treat quantum ML as research unless the claim names a task, a data-access model, a baseline, and a hardware assumption. Four specifics, or it is a demo.
Quantum computing touches cryptography in two very different ways, and keeping them separate is half the battle.
The Shor risk
A large fault-tolerant quantum computer would break RSA and elliptic-curve cryptography. The near-term action is migration to post-quantum cryptography; a current enterprise and government security task, not something to schedule for when the hardware arrives. Data intercepted now can be decrypted later, which is what makes waiting expensive.
QKD
Quantum key distribution uses quantum states plus an authenticated classical channel to detect eavesdropping, under specific assumptions about devices and links. It is infrastructure-specific; point-to-point links with dedicated hardware; and not a universal replacement for PQC, which runs in software everywhere.
Strategic view: PQC migration is the broad, near-term software and security market; QKD serves particular high-value links.
Evidence boundary. Near-term optimization and machine-learning claims require strong benchmark discipline and do not inherit advantage from quantum vocabulary. [John Preskill] [Marco Cerezo et al.] [Edward Farhi]
Notation contract: Use E0–E5 only as ordinal evidence labels, never arithmetic quantities; dates use ISO 8601; distinguish quantum-computing and quantum-sensing categories.
An evidence ladder with expiry dates
Grade narrative, theory, synthetic tests, realistic tests, pilots, and outcomes.
Quantum sensing can matter commercially before general quantum computing because it does not require universal fault-tolerant computation. Applications include inertial navigation, magnetic sensing, timing, gravity sensing, defense and aerospace, geophysics, and medical or materials measurement.
Strategic view: do not collapse quantum sensing into quantum computing. Different products, timelines, buyers, and proof standards; and conflating them is a common way to borrow credibility from one for the other.
| Decision | Use when |
|---|---|
| Build | You have a specific wedge, proprietary distribution or data, and a credible technical path |
| Partner | The value lives in your domain workflow and quantum is an enabling component |
| Invest | The team has technical evidence, differentiated timing, and a plausible moat |
| Wait | The opportunity is real but depends on hardware milestones outside your control |
| Avoid | The claim lacks a baseline, evidence, or buyer urgency |
"Wait" is the underrated row. Monitoring a real opportunity while its dependencies mature is a strategy; pretending the dependencies are already met is not.
Five actions and their reversal conditions
Turn evidence and integration cost into an accountable decision.
Nearer-term company opportunities tend to cluster around the ecosystem rather than inside universal quantum advantage:
- developer education and labs
- resource-estimation tooling
- hardware-aware compilation
- error-correction tooling and visualization
- PQC migration and cryptographic inventory
- domain-specific simulation workflows
- benchmarking and diligence platforms
- quantum/HPC workflow orchestration
- sensing applications
- control and calibration tooling
Notice what these have in common: each sells something the field needs before fault tolerance arrives. Selling picks during a gold rush is a cliché because it keeps working.
- Rank chemistry, optimization, quantum ML, PQC, and sensing by near-term commercial readiness, and defend the ordering.
- Choose one company and classify its public claims by rung of the evidence ladder.
- Write a build-partner-invest-wait-avoid memo for one application.
- Identify a quantum-adjacent wedge that does not require fault-tolerant hardware.
- Compare QKD and PQC as answers to enterprise security migration.
- Write the strongest skeptical case against your favorite quantum application.
Application evidence dossier template
| Field | Reader-visible record |
|---|---|
| Format | Versioned JSON schema plus two fully sourced example dossiers |
| Verification | Schema validation requires workflow owner, baseline, dates, evidence level, integration boundary, decision, and reversal condition; examples reproduce their scores. |
| Availability | Source-embedded acceptance record; no separate download is claimed |
{
"artifact": "Application evidence dossier template",
"format": "Versioned JSON schema plus two fully sourced example dossiers",
"acceptance_test": "Schema validation requires workflow owner, baseline, dates, evidence level, integration boundary, decision, and reversal condition; examples reproduce their scores.",
"publication_state": "source-embedded contract and worked fixture"
}Scope boundary
- This guide does not predict market size or treat adjacent quantum sensing as proof of quantum-computing value.
- It does not convert research novelty, small synthetic benchmarks, or vendor access into customer ROI.
Depth commitment. One workflow decomposition, one evidence ladder, two dossiers, and five action definitions.
Practice problem
Grade two application claims—one chemistry workflow and one routing optimizer—using the evidence ladder and choose an action for each.
| Claim | Record supplied |
|---|---|
| Chemistry: lower ground-state energy for a four-orbital active space | Instance, code, exact-diagonalization baseline, five seeds, and complete energy table. |
| Routing: 12% lower cost on customer data | Vendor summary with no dataset, integration cost, or named classical optimizer. |
- Deliverable
- Two source-backed dossier records with explicit reversal conditions.
- Pass condition
- The model solution shows why similar evidence labels can yield different actions when integration cost and baselines differ.
Verification record
Expected solution form. Completed dossiers plus a rubric that separates evidence strength from decision readiness.
Model answer. The chemistry record reaches reproducible small-instance evidence but still needs scale and workflow evidence; a bounded experiment is warranted. The routing statement remains an unsupported comparative claim because neither the instance nor baseline can be inspected; the action is wait or request evidence, not build.
Model result and check. A schema checker validates required fields; a sensitivity worksheet reproduces each action under stated weights.
Acceptance test. The model solution shows why similar evidence labels can yield different actions when integration cost and baselines differ.
Provenance
Sources and review
- National Academies of Sciences, Engineering, and Medicine. Quantum Computing: Progress and Prospects. National Academies Press. 2019consensus study report
- U.S. Government Accountability Office. Quantum Computing and Communications: Status and Prospects. GAO. 2021government technology assessment
- Yudong Cao et al.. Quantum Chemistry in the Age of Quantum Computing. Chemical Reviews. 2019peer-reviewed review
- John Preskill. Quantum Computing in the NISQ era and beyond. Quantum. 2018peer-reviewed perspective
- Marco Cerezo et al.. Challenges and opportunities in quantum machine learning. Nature Computational Science. 2022peer-reviewed perspective
- Edward Farhi, Jeffrey Goldstone, and Sam Gutmann. A Quantum Approximate Optimization Algorithm. arXiv. 2014primary preprint
The load-bearing claims in the chapter are mapped inline to this registered source set. A citation supports only the bounded claim beside it.