JOIN MFTA
JOIN MFTA

du Launches National HyperAI Sovereign AI Platform Across Five UAE Data Centres

Rows of illuminated server racks in a modern data centre aisle with overhead cabling.

du, the Dubai-based telecom and digital services provider, has introduced National HyperAI, a platform for sovereign AI inference and agentic workloads. The company said the service is available immediately and is aimed at UAE government bodies, large companies and smaller businesses. Its infrastructure is spread across five UAE data centres, which du presents as a way of keeping sensitive information within national borders.

What the platform bundles

The service is operated by du Tech and brings three elements together under a single commercial agreement: GPU-as-a-Service for raw compute, Model-as-a-Service for access to hosted models, and autonomous agent workflows that carry out tasks with limited step-by-step human direction. Under one contract, a customer can therefore draw on all three layers rather than arranging each separately.

According to du, the full stack includes the following components.

  • Hyperscale data centres with liquid cooling
  • National fibre networks and 5.5G connectivity
  • Bare-metal NVIDIA GPU compute
  • MLOps orchestration for managing models through their lifecycle
  • Curated open-weight models
  • Agentic orchestration built with ecosystem partners

Why in-country hosting matters

Sovereign AI generally refers to running AI systems on infrastructure that stays under national control, so that data and computing resources remain within a country's borders. du said that under National HyperAI, customer data, model weights and computing capacity all stay inside the UAE. Model weights are the learned parameters that shape how an AI model responds, which is why their location is a separate point from the location of the data itself.

Data residency requirements are a common concern for public bodies and regulated firms that handle personal or sensitive records. For such organisations, the question is often not only where information is stored but also where it is processed when an AI model answers a query. du's description places both steps inside the UAE.

From pilots to production

AI inference is the stage at which a trained model takes new inputs and produces outputs, and it is the part of an AI system that runs continuously once deployed. The company said National HyperAI is meant to help organisations shift AI projects out of the pilot phase and into production without buying their own specialised hardware.

du added that customers can attach critical workloads, sovereign data stores and business applications to inference calls, and still keep their existing compliance and governance measures in place. Maintaining those controls is a practical requirement for organisations under strict regulatory oversight, and the company framed this as a core part of the offering.

Who the service is for

du said the platform targets organisations that have regulatory, security and operational requirements. The buyers it named include ministries of government, national security bodies, big corporates and small and medium enterprises. Serving both public-sector and commercial customers from one platform is the company's stated approach.

The platform is built on du Tech's National Hypercloud. GPU compute refers to graphics processors that handle the heavy mathematical work of AI models, while bare-metal means dedicated physical servers rather than shared virtual machines. The service is described as operating within the UAE, and the announcement does not describe plans beyond that.

Fireblocks: The Financial Grid Middle East  – (Vertical)
Intellect – (Vertical)
Fimple – BaaS Solution (Vertical)
Fireblocks: The Financial Grid Middle East  – (Square)
Intellect – (Square)
Fimple – Website (Square)

Events & Webinars

MFTA Reports

Relevant News

Recent Webinars