Confidential quantum optimization

The quantum vendor sees mathematics. Never what it means.

ArcaQ runs portfolio, scheduling, routing, and ground-state optimization on real quantum hardware — while your business context is decrypted only inside an attested confidential enclave. Only the abstracted mathematics crosses to the quantum processor. Math leaves the enclave; meaning never does.

Request accessHow it works

ArcaQ is the newest product on the ArcaKey confidential-compute platform, already in production across healthcare, legal, and finance. Confidential by architecture, not by promise.


The architecture

A quantum computer has to observe the circuit it runs. That is physics, not a vendor choice — so every commercial quantum-optimization service today shows the hardware vendor what you are optimizing. For institutional capital, the circuit is the asset: strategy parameters, position sheets, beneficiary allocations.

ArcaQ removes that compromise structurally. Your natural-language problem reaches the ArcaQ enclave over a post-quantum envelope; the edge tier forwards ciphertext and holds no key. Inside attested confidential compute, an in-enclave language model translates the problem into a Quadratic Unconstrained Binary Optimization (QUBO) matrix. Only that abstracted mathematical object — coefficients and a gate sequence — crosses the boundary to the quantum processor. The result returns by the same path and is interpreted back into business meaning using context that never left the enclave.

A 64-by-64 matrix of coefficients could encode a portfolio rebalance, a supply-chain route, or a molecular configuration — the mathematics is identical. The business meaning lives in how those coefficients were derived, and that derivation happens entirely inside attested compute. The vendor sees a circuit; it cannot reverse-engineer what the circuit represents.

Math leaves the enclave. Meaning never does.


How it works

Every problem runs through a template — portfolio rebalance, tax-aware allocation, M&A screening, real-estate mix, and more — with three ways to provide inputs: CSV upload, a structured form, or natural language. From there, two execution paths: validate on the simulator, then run on real quantum hardware when you're ready.

Simulate

Run on the in-enclave simulator — unlimited iteration and fast feedback on parameter quality before you ever touch real hardware.

Run on quantum hardware

Every dispatch to real quantum hardware waits for your explicit confirmation — a run never happens without your go-ahead. Results return as a plain-English explanation first, with the raw mathematics one click away.

Behind that simple surface, every step of the request lifecycle is recorded with a post-quantum signature bound to attestation evidence:

  1. 01

    Sealed submission

    Your problem is sealed in the browser under a post-quantum envelope and forwarded as ciphertext. It is decrypted only inside the attested enclave.

  2. 02

    In-enclave translation

    An attested language model translates the problem into a QUBO matrix and circuit — entirely inside confidential compute. Reserve and Grand Reserve route this through Confidential Council for higher-confidence translation.

  3. 03

    Your confirmation

    Every dispatch to real quantum hardware waits for your explicit confirmation. Simulator runs need none — iterate as much as you like.

  4. 04

    Mathematics only crosses the boundary

    The QUBO coefficients and gate sequence — and nothing else from your request — leave the enclave to the quantum processor at IonQ or Rigetti.

  5. 05

    Interpretation and signed delivery

    The result returns to the enclave, is interpreted in the context of your original problem (which never left), and is delivered over your sealed session. The audit entry closes with a signature binding every prior step.


Verifiable, not promised

Confidentiality that comes from architecture, not from a vendor's policy. Each guarantee is something you can check, not something you have to trust.

Attested execution

Your business context is decrypted only inside a hardware-isolated enclave. You verify the environment cryptographically before you send anything.

GCP Confidential Space · Intel TDX
Post-quantum transport

The envelope between your browser and the enclave uses NIST-standardized post-quantum key encapsulation and signatures — protection built to outlast the arrival of quantum computers.

ML-KEM-768 (FIPS 203) · ML-DSA-65 (FIPS 204) · AES-256-GCM
Signed audit chain

Every step of every run is recorded twice — a fast classical signature and a post-quantum signature — bound to attestation evidence and verifiable offline against published keys.

Ed25519 + ML-DSA-65
No one sees your data

No ArcaKey staff member sees plaintext customer problems — including in Concierge advisory engagements. The quantum vendor sees mathematics, not meaning.

Enclave-sealed end to end

Tiers

Three tiers, plus an optional Concierge advisory overlay. Annual billing is the headline; Atelier and Reserve also offer a monthly option for single-cycle exploration.

Atelier
$9,999/yr
or $999/mo

The working tier — everything for episodic optimization, self-serve.

6 hardware runs / year · Single-model translation
  • Unlimited local simulator iteration
  • Standard template library (portfolio rebalance, tax-aware allocation, M&A screening, real-estate mix, and more)
  • Three inputs per template: CSV upload, structured form, or natural language
  • Per-session attestation panel — verify the enclave before you send
  • Post-quantum transport (ML-KEM-768 / ML-DSA-65) and signed audit chain
  • Fully self-serve — no human ever sees your data
Request access
Reserve
$19,999/yr
or $1,999/mo

The practiced tier — structural upgrades for recurring use.

12 hardware runs / year · Council-augmented translation (always-on)
  • Everything in Atelier
  • Council-augmented translation always on — three attested models plus meta-summarizer for higher-confidence QUBO translation
  • Reserve-only advanced templates (multi-objective, time-series-aware, hierarchical constructions)
  • QPU slot reservation for time-sensitive runs
  • One annual onboarding session with an ArcaKey engineer
  • Quarterly cryptographically-signed activity reports
Request access
Grand Reserve
from $99,000/yr

The institutional tier — dedicated, region-pinned, supported.

Unlimited hardware runs · Council-augmented translation (always-on)
  • Everything in Reserve
  • Dedicated enclave provisioned to your specification — optional region pinning for data residency (deployable in your jurisdiction)
  • Unlimited quantum hardware runs; $5,000 included Braket credit; customer-configurable caps
  • Named engineer with a 1-hour SLA during business hours
  • Custom template development (designed without ever seeing your data)
  • Monthly signed activity reports
  • Concierge service overlay included at no additional cost
Request access
  • A “run” is a dispatch to real quantum hardware. Local simulator iteration is unlimited at every tier.
  • ArcaKey Executive Vault customers can add ArcaQ Atelier at $7,000/yr — about 30% off standalone — and upgrade to Reserve or Grand Reserve at standalone pricing.
  • Concierge advisory overlay: $5,000/yr, or $1,500–3,000 per session. Included at no cost with Grand Reserve. No ArcaKey staff member ever sees your data.

One platform, not a point tool

ArcaQ is one product on the ArcaKey confidential-compute platform. The same architecture — attested hardware, post-quantum transport, signed audit — runs your confidential AI workspace (ArcaKey AI), multi-model attested deliberation (Confidential Council), continuous compliance monitoring (CDM), and clinical dictation (ArcaVoice).

An organization that adopts ArcaQ is, by definition, the buyer for confidential AI and signed compliance evidence too. CDM can monitor your ArcaQ activity alongside your AI Vault on a single signed report — one defensibility surface across everything you run.


For institutional capital

ArcaQ is the family's highest-trust entry point for the most security-conscious buyers — quant funds, sovereign wealth, family offices. Grand Reserve deployments run on a dedicated, region-pinned enclave: deployable in your jurisdiction, the direct answer to data-residency requirements.


Request access

ArcaQ access opens by request. Tell us the problem class you optimize and the confidentiality requirements you operate under, and we will respond with tier fit and next steps. Reserve and Grand Reserve include a guided onboarding.

Request access

Confidential by default. Mutual NDA available before any technical disclosure.

ArcaQ performs combinatorial optimization on today's noisy intermediate-scale quantum hardware. It is not investment advice and executes no trades. Performance ranges shown on use-case pages are drawn from published academic and industry benchmarks for the relevant problem class, not from ArcaQ measurements; ArcaQ-specific results will be published after hardware validation. The confidentiality claim is precise: end-user business context is decrypted only inside attested compute, and the data leaving toward the quantum vendor is mathematical, not contextual. ArcaQ does not claim full quantum homomorphism or attestation of the quantum processor itself, neither of which is achievable on any platform today.

ArcaQ — Confidential Quantum Optimization