Sovereign AI, priced like capacity you control — never metered by the token.
Iftah installs a complete governed-access and governance layer inside your own Kubernetes cluster, in your own region — prompts, inference, model weights, and audit data never leave that boundary. We don't host your data; we operate your install remotely under a time-bound, client-approved, fully audited break-glass key. So we price the platform, not your usage: a fixed-fee 90-day pilot to de-risk, then a committed annual subscription sized to the governed capacity you already own. No per-token meter, no phone-home — and the GPUs and cloud you run on stay yours.
How pricing works
- Capacity-based, not consumption-based — license governed capacity per year, never per token or request
- No telemetry callback — the control plane carries health, version, and latency metrics only, never content
- Four costs, named plainly — platform, deployment, operations, and your own infrastructure (not Iftah revenue)
- Air-gap honest — offline license enforcement means pricing works with zero egress to meter against
How Iftah is priced — four costs, fully separated.
We charge for the governance platform and its remote, audited operation — not for the silicon you already own, and not for tokens we cannot and will not see. Procurement can budget all four lines before the first technical call.
Platform subscription
An annual, capacity-based license for the full single-tenant governed stack — all six products on one install. Sized by governed accelerator-node band or as a flat per-cluster site license; 3- and 5-year terms lower the effective rate. Enforced offline — no phone-home, no usage meter inside your perimeter.
Deployment & hardening
The fixed-fee 90-day pilot: we install into your cluster, integrate your identity provider via a client-local key mirror, tune guardrails for Arabic and Gulf identifiers, and drive the production cutover gate-set. Credited toward your production contract on cutover.
Operations & support
An annual contract for remote, break-glass operation under a written RACI — signed GitOps releases, fleet patch and CVE management, and SLA-backed support. Standard support is included; 24/7 mission-critical response and a named lead are priced as upgrades.
Your GPUs & infrastructure
You pay your cloud provider or hardware vendor directly for compute, storage, and accelerators — in your account, in your region. We never resell, mark up, or meter it. Stated plainly so total cost of ownership is honest from day one.
From a fixed-fee pilot to fully sovereign production.
One spine: prove it, run it, then sovereign-harden it. Every tier ships the full governance floor — deployment mode and operations depth are priced on top, never the entry.
Pilot
Prove governed AI on your own infrastructure — before committing budget.
How it's billed
Fixed-fee · 90 days
One number, no usage variability — credited toward production on cutover. Design Partners receive reduced founding-customer pricing.
- Single-tenant install into your cluster and region — synthetic / non-PHI traffic during the pilot
- Federated SSO via a client-local key mirror; default-deny authorization on every call
- Self-hosted guardrails (Presidio PII/PHI, NeMo injection/jailbreak), tuned for Arabic and Gulf identifiers
- Drives the production cutover gate-set with signed evidence
- Bring your own open models, including Arabic-capable (Falcon, Jais, ALLaM) — no per-model fee
- Fixed scope, fixed price — a de-risking entry point, not a feature-starved trial
Production
Live production on cloud, private, or sovereign-regional cloud — real users and data, one region.
How it's billed
Committed annual · 3 & 5-yr terms
Operations & support contracted alongside; air-gap, 24/7, and a multi-cloud control plane are priced as upgrades.
- The full governed stack — all six products: Deploy, Gateway, Observe, Gatekeeper, AI Service, ClusterGuard
- Digest-pinned model routing and signed GitOps releases — even Iftah cannot hand-edit a running install
- Immutable in-region WORM audit trail, fail-closed — no audit, no request, never a silent gap
- Remote break-glass operation under a written RACI — time-boxed, client-approved, logged in your own trail
- Per-data-class RPO/RTO targets, multi-AZ high availability, in-region backup and restore
- RAG, fine-tuning (LoRA/QLoRA), and agents — all inside your boundary, no content egress
Sovereign
On-prem or fully air-gapped for government, defense, central banks, and critical infrastructure.
How it's billed
Custom annual / multi-year
A capacity site license plus a dedicated managed-operations fee, billed as separate line items.
- Fully air-gapped install — offline license enforcement, signed offline update bundles, zero egress
- Portable WORM and secrets substrate for on-prem and OpenShift, in-region only
- Dedicated operations: 24/7 mission-critical response, named lead, accelerated RPO/RTO
- Department-level isolation and the strictest residency, clearance, and approval controls
- Optional multi-cloud / hybrid central control plane — metrics-only, never content
- Configured per enclave — drivers and structure published so procurement can budget up front
Four drivers. No surprises.
Pricing tracks the governance value you deploy and the depth at which we operate it — not your token usage.
Governed capacity in your cluster
Priced by accelerator-node band — or a flat per-cluster site license — not per physical GPU, so dense, partitioned, or idle hardware is never punitively taxed. You scale capacity you control; the license tracks governance value, not silicon.
Deployment mode
Cloud VPC and private / regional sovereign cloud sit at the base; on-prem and fully air-gapped carry a sovereignty premium for the heavier assurance, offline enforcement, and operations they demand.
Term length
Annual is the floor; 3- and 5-year committed terms lower the effective rate, rewarding the budget-stable horizon sovereign and government buyers prefer. Committed-capacity bands lower it further.
Operations & support depth
Standard support is included in the subscription. Business-critical 24/7 response, a named operations lead, and accelerated per-data-class RPO/RTO are premium upgrades — never baked into the entry price.
The sovereignty floor is never an upgrade.
Every install — pilot to sovereign — ships the same governance guarantees. These are standard, not paywalled.
- Identity & policy on every call
Federated SSO and service-account identity via a client-local key mirror — no per-request egress to your IdP — with policy-as-code making a default-deny, fail-closed authorization decision on every request.
- Self-hosted guardrails, tuned for Arabic
Presidio PII/PHI detection and NeMo injection/jailbreak rails run in-region as local pods — never a SaaS guardrail API — tuned for Arabic-script and Gulf-specific identifiers. PHI handling is fail-closed by default.
- Immutable in-region WORM audit
Every governance decision and outcome is written to a tamper-evident, append-only, in-region store with a partitioned hash-chain. If the audit sink is unavailable the platform fails closed to a durable buffer — never a silent gap.
- Break-glass operation, never standing access
Iftah operates your install under a time-bound, client-approved, fully audited break-glass key governed by a written RACI. We hold no standing access to your prompts, responses, or pods; all changes flow as signed GitOps artifacts.
Answers for your procurement and security teams.
The questions regulated buyers ask first.
Next step