announced a strategic investment and partnership with Arabic.ai. 
Private cloud
A dedicated environment with agreed tenancy, location and administration.
Deployment options
Choose where the intelligence runs.
Private cloud, customer-controlled, on-premises or disconnected. Each is scoped with your team.
Four ways to deploy.
From a dedicated cloud to hardware with no outbound connection.
Customer-controlled
You administer infrastructure, access and change, in a cloud or on your premises.
On-premises
An agreed configuration on your infrastructure, sized together.
Disconnected (air-gap)
Fully disconnected inference on your hardware, with no outbound connections and no cloud, for government, defense and data that cannot leave the building.
Data residency in the Gulf.
Residency in the UAE, Saudi Arabia, Kuwait and Qatar.
- United Arab Emirates
- Saudi Arabia
- Kuwait
- Qatar
By product.
| Product | Starts as | Confirm together |
|---|---|---|
  | Scoped to the selected model | Compute, model version, disconnected requirements |
 | Public application or enterprise discussion | Hosting, administration, information access |
 | Dedicated cloud or on-premises | Connected tools, approval steps, operating responsibility |
 | A scoping call, then a pilot | The task, success criteria and residency in MENA or the EU |
Sovereign legal AI, on premises.
Arabic.ai and Qistas combine LLM-X, LLM-S and on-premises deployment with Qistas's legal technology for research, litigation, transactions, and policy and legislative studies.
See our partners
Plan for real load.
Workload
Task types, document sizes, languages, concurrency and peaks, including the review queue.
Business owner
Performance
Test realistic inputs against agreed quality and response requirements.
Technical team
Recovery
Decide what can be retried, what needs reconciliation and who approves resumption.
Service owner
Support
Set responsibilities for triage, change and review.
Customer operations
Questions to settle early.
Can it run without an internet connection?
Yes. A disconnected (air-gap) deployment runs inference on your hardware, with no outbound connections and no cloud.
Where can our data stay?
In the UAE, Saudi Arabia, Kuwait or Qatar. From November 2026, Arabic.ai solutions also run in Microsoft's Saudi Arabia East region, with in-Kingdom data residency.
Which products run on premises?
LLM Models, Speech Models, Atelier and Sadu can each be deployed in a dedicated cloud or on premises.
Who runs updates and support?
That is agreed for each deployment: who runs infrastructure, updates and support, and who owns triage, change and review.
How is an on-premises deployment sized?
Together, from your workload: task types, document sizes, languages, concurrency and peaks, including the review queue.
How is the model version handled?
It is agreed together, with compute and any disconnected requirements. You know which model is running, its configuration and how changes are reviewed.
Make your next ambition operational.
Bring intelligence into your knowledge, people and operations. Talk to our team about the work you want to transform.