Part IX. Company and Investing · Chapter 74
Hardware Modality Diligence
Superconducting, trapped-ion, neutral-atom, photonic, silicon-spin — every hardware pitch opens with a modality, and every modality sounds inevitable from the inside. This chapter gives you a way to compare them: a matrix of physical facts, weak layers, and proof gates that turns a hardware pitch into a scoreable memo.
In this chapter 9 sections
Compare modalities through a workload-specific resource translation that preserves native operations, connectivity, cycle time, measurement, error model, control footprint, and demonstrated scale; raw qubit count is only one input.
Hardware modalities expose different physical nouns, which makes a single comparison column dangerous. A qubit can be an anharmonic circuit, ionic internal state, atomic level, photon, electron spin, oscillator code, or proposed topological degree of freedom. Gate, connectivity, time, readout, leakage, control, and manufacturing mean different things in each system.
Translate a workload before comparing devices
A comparison begins with a fixed workload and compiles it separately. The matrix records encoding, native entangler, physical connectivity, logical-to-native expansion, two-qubit duration, fidelity-estimation method, readout, reset, parallelism, control footprint, and demonstrated system scale. Cells remain not comparable when method or boundary differs.
- Workload
- Logical operations, connectivity, and timing required.
- Translation
- Routing, decomposition, heralding, and control overhead.
- Observation
- Task quality with uncertainty and a complete resource boundary.
Modality matrix with comparability flags
| Claim under review | Source chain |
|---|---|
| Physical implementations differ in encoding, control, native interactions, measurement, topology, and dominant error channels. | |
| Randomized benchmarking estimates average error under a protocol and does not make every device or workload directly comparable. | |
| Bosonic and oscillator encodings change the mapping from physical modes to protected logical information. | |
| Open instruction semantics and reproducibility records expose some, but not all, cross-stack translation assumptions. |
A Bell-pair workload changes under three native entanglers
| Modality record | Native interaction | Conversion boundary | Comparability |
|---|---|---|---|
| Superconducting | nearest-neighbor entangler | routing and leakage | pending matched run |
| Trapped ion | collective entangler | duration and mode crowding | pending matched run |
| Photonics | measurement-mediated | success and loss convention | pending matched run |
| Oscillator code | mode-level operations | encoding overhead | pending matched run |
A Bell-pair workload between two logical endpoints may use one native entangler on an all-to-all ion device, a routed sequence on a nearest-neighbor superconducting layout, or measurement-mediated photonic operations with a different success convention. Reporting one number called two-qubit fidelity would erase compilation, heralding, timing, and sampling differences. The matrix marks those cells incompatible until a task-level result is measured.
FIELDS = {"workload": str, "topology": str, "gate_duration": dict, "readout": str,
"fidelity_method": str, "conversion": dict, "source_id": str}
def compare(matrix, maximum_compiled_duration_ns):
errors = []
durations = {}
for name, row in matrix.items():
missing = [field for field, kind in FIELDS.items()
if field not in row or type(row[field]) is not kind or not row[field]]
gate, conversion = row.get("gate_duration", {}), row.get("conversion", {})
if type(gate.get("value")) not in (int, float) or gate.get("unit") != "ns":
missing.append("gate_duration_schema")
if type(conversion.get("count")) is not int or conversion.get("unit") != "two-qubit gates":
missing.append("conversion_schema")
if missing:
errors.append(name + ":" + ",".join(sorted(set(missing))))
else:
durations[name] = gate["value"] * conversion["count"]
eligible = {name: value for name, value in durations.items()
if value <= maximum_compiled_duration_ns}
selected = min(eligible, key=eligible.get) if eligible else None
return {"decision": selected or "not-comparable", "errors": errors,
"compiled_duration_ns": durations, "eligible": sorted(eligible)}
matrix = {
"device-a": {"workload": "Bell pair endpoints 0-2", "topology": "line",
"gate_duration": {"value": 250, "unit": "ns"}, "readout": "joint Z samples",
"fidelity_method": "interleaved RB", "conversion": {"count": 4, "unit": "two-qubit gates"},
"source_id": "scenario:device-a calibration"},
"device-b": {"workload": "Bell pair endpoints 0-2", "topology": "all-to-all",
"gate_duration": {"value": 600, "unit": "ns"}, "readout": "joint Z samples",
"fidelity_method": "cycle benchmark", "source_id": "scenario:device-b calibration"},
"device-c": {"workload": "Bell pair endpoints 0-2", "topology": "photonic graph",
"gate_duration": {"value": 80, "unit": "ns"}, "readout": "heralded samples",
"conversion": {"count": 8, "unit": "two-qubit gates"},
"source_id": "scenario:device-c calibration"}}
base_result = compare(matrix, 1200)
bad_a = {**matrix["device-a"], "gate_duration": {"value": 250, "unit": ""}}
bad_result = compare({**matrix, "device-a": bad_a}, 1200)
strict_result = compare(matrix, 900)
assert base_result["decision"] == "device-a" and base_result["eligible"] == ["device-a"]
assert bad_result["decision"] == "not-comparable" and "gate_duration_schema" in bad_result["errors"][0]
assert strict_result["decision"] == "not-comparable"
print(f"PASS: 74 modality matrix selected={base_result['decision']} duration={base_result['compiled_duration_ns']} invalid={bad_result['errors']} budget900ns={strict_result['decision']}")Artifact contract. A dated units-bearing table plus executable comparability checker. Rows cannot be ranked unless workload, fidelity method, topology, gate duration, readout, and conversion overhead are all comparable; missing fields produce 'not comparable'.
Exact validation command: python3 tools/validate_briefs.py --briefs data/editorial_briefs_64_87.json --from 64 --through 87 --check-rewritten-sources --execute-artifacts
Control and packaging enter the denominator
Select only for the declared workload and procurement constraint. One device may minimize wall-clock latency, another control footprint, another logical error under a future code. A defensible recommendation shows the conversion ledger and the fields it could not compare. The absence of a total winner is often the technically correct output.
Audit a modality comparison for hidden conversions
Prompt. Compare three modalities for a fixed entanglement-and-measurement workload.
Deliverable. Matrix with encoding, native entangler, connectivity, compiled operations, duration, error evidence, readout, control footprint, and comparability flags.
Pass condition. All unit conversions are shown, incomparable cells remain unlabeled rather than scored, and the recommendation is explicitly workload-bound.
Model answer: no universal winner, one workload fit
Format. Three-row modality matrix that yields one workload fit but no total ranking.
The model answer identifies one workload fit after filling compilation and timing fields, while leaving a universal modality ranking blank. It notes that randomized-benchmarking numbers came from different methods and conditions and therefore cannot be ordered directly. The executable check returns comparable only for rows containing the complete workload, topology, timing, readout, and method record.
Verification. The checker prevents ranking until all required comparability fields exist and reproduces the workload-specific selection from the completed rows.
Companion work
Artifacts for this chapter
These entries resolve to checked-in local source. Commands are reproduced exactly from the chapter manifest, and source-embedded fixtures are exported as direct downloads.
diligence dossier
Workload-normalized modality matrix
Reproduce or test
python3 tools/validate_briefs.py --briefs data/editorial_briefs_64_87.json --from 64 --through 87 --check-rewritten-sources --execute-artifacts
Provenance
Sources and review
- U.S. Government Accountability Office. Quantum Computing and Communications: Status and Prospects. GAO. 2021government technology assessment
- Easwar Magesan, J. M. Gambetta, and Joseph Emerson. Scalable and robust randomized benchmarking of quantum processes. Physical Review Letters. 2011primary paper
- Association for Computing Machinery. Artifact Review and Badging. ACM Publications. 2026official reproducibility policy
- OpenQASM Technical Steering Committee. OpenQASM 3 specification. Linux Foundation Joint Development Foundation. 2026official technical specification
- Daniel Gottesman, Alexei Kitaev, and John Preskill. Encoding a qubit in an oscillator. Physical Review A. 2001primary paper
- T. D. Ladd et al.. Quantum computers. Nature. 2010peer-reviewed review
- Lieven M. K. Vandersypen et al.. A look at the full stack. Nature Reviews Physics. 2021peer-reviewed perspective
- Timothy Proctor et al.. Benchmarking quantum computers. Nature Reviews Physics. 2025peer-reviewed perspective
The load-bearing claims in the chapter are mapped inline to this registered source set. A citation supports only the bounded claim beside it.