Skip to content

Pricing

Two editions. Both run on your hardware.

Enterprise licenses the platform to one organisation; Service Provider licenses it to a provider serving many. Either way we supply the software, the migration and third-line support. Your own team or your provider runs it.

For one organisation

ATAILA Enterprise

Private AI and app delivery on your own servers or in a private datacentre.

Where
On-premises, or in your private datacentre.
Tenancy
Single-tenant, one organisation. Running it for third parties is not licensed.
Included
All six capabilities.
From ATAILA
The software, the migration and third-line support. Your team runs it.

Talk to us about Enterprise

For a service provider serving its own customers

ATAILA Service Provider

Offer private AI and app platforms to your own customers, in your own datacentre.

Where
Your own datacentre, serving your own customers.
Tenancy
Multi-tenant, with hosting and resale rights.
Included
Everything in Enterprise, plus tenant management, per-tenant usage data for billing and a migration path for Windows Azure Pack estates.
Branding
Co-branded by default, “powered by ATAILA”. White-label is an add-on module.
From ATAILA
The software, the migration and third-line support. Your team runs it.

See the Service Provider edition

Included in both editions

All six capabilities come with either edition. The platform page explains what each one does; this list says what the subscription covers for it.

  1. 01. Load models

    The GPU nodes you load models onto are covered by the socket subscription.

  2. 02. Model library

    Nothing is charged per GPU, however many models you keep.

  3. 03. Metering and audit

    The Service Provider edition adds usage data per tenant, for your own billing.

  4. 04. AI developer workstations

    No per-seat or per-user charge for the developers who use them.

  5. 05. Auditable RAG

    No per-token charge, however many tokens your documents put through the models.

  6. 06. Project factory

    No per-VM charge for anything it creates.
See the platform in detail

How pricing works

Four parts make up every offer: the subscription, the implementation, support and the quote itself.

Per physical CPU socket
A flat annual subscription per physical CPU socket, on every node type, GPU nodes included. Never per user, seat, token, GPU or VM.
A short implementation
A one-to-four-week implementation, billed weekly.
Support in packages
Third-line support, updates and security fixes in packages. Your team or your service provider runs the platform day to day.
Quoted per estate
Quoted in writing after we have seen your estate, usually within days. The offer is itemised and commits you to nothing.

Questions buyers ask

Where does it run?
On your own servers or in your private datacentre; with the Service Provider edition, in the provider’s datacentre. Your data and your workloads stay inside the jurisdiction you chose.
What do we own, and can we leave?
Everything you build. The apps live in your own repositories, the models are open and yours to keep, and the data stays wherever you run the platform. You can leave at any time, with a documented export on demand and an exit written into the contract before you need it.
What happens if ATAILA stops trading?
The exit guarantee applies. An official copyright deposit plus a written commitment, agreed in the contract. Your apps, models and data are yours regardless.
Can we run Enterprise for our own customers?
No. Enterprise is single-tenant and licensed to one organisation. Running the platform for third parties takes the Service Provider edition, which carries the hosting and resale rights, tenant management and per-tenant usage data for billing.
What hardware do we need?
Standard servers from more than one vendor, including your existing fleet, with GPU servers for the AI work. We keep the list of supported server models and confirm yours before you buy anything.
What support do we get?
You get third-line support from ATAILA: when your team or your service provider needs to escalate, it reaches the engineers who build the platform. Updates and security fixes come from us too. Running the platform day to day stays with your team or your provider. Support comes in packages, quoted together with the subscription.
Is per-token ever the better deal?
Yes. If your team touches AI a few times a week, a public per-token service is cheaper and we will tell you so. A private platform pays off when the work is confidential, continuous, or both.
Why are there no prices on this page?
Because the number depends on your estate, and it arrives in the written offer after the first conversation.

Want a number for your estate?

Tell us what you run and what you want to put on it. You get a written, itemised offer.

Talk to us