Since July 2026, advanced NVIDIA GPUs go to the UAE only to entities approved under the US framework or under a US export licence. We supply B300, GB300 NVL72 and H200 to projects that qualify, and handle the documentation through our Dubai free-zone hub.
Approval belongs to the end user and the project, not to a supplier's stock. We start with the documentation that decides it — who will run the hardware, where, and for what.
| Platform | Spec | How it is supplied |
|---|---|---|
| HGX B300 | 8 × B300 288 GB, ~14.5 kW; liquid at cluster density | Allocation to documented projects |
| GB300 NVL72 | 72 × B300 in one liquid-cooled rack, ~132 kW | Project-quoted as a complete rack |
| HGX H200 | 8 × H200 141 GB, ~10.2 kW; air-coolable | Project order; lower entry cost |
| H200 NVL nodes | 2–4 × H200 NVL PCIe for inference | Project order for approved end users |
| HGX B200, H100 | Previous platforms | On request |
| Server platforms | Supermicro, Dell, HPE, Lenovo | Chosen per project |
| Fabric | 800G / 400G InfiniBand, Spectrum-X, ConnectX | Planned with the cluster |
| DGX Spark, RTX PRO 6000 | Desk-side and workstation AI | Usually available |
How a project runs: our procurement process → · What to prepare: what we ask before a supply request →
Under the US rule in force since July 2026, UAE government entities, specifically approved UAE AI companies and subsidiaries of certain US hyperscalers can receive them without a licence; other end users need a US export licence. We check which applies before quoting.
Data-center GPUs are supplied to approved projects, not held as open stock. Desk-side and workstation GPUs are available on request.
Yes, to projects that clear the export rules for those destinations, with end-user screening, certificates of origin and destination customs handled.
End-user documentation and supplier review take 1–2 weeks. Where a licence is needed, its timeline is set by the authority; we plan the project around it.
Running AI on this infrastructure? Haink also builds the LLM & ML software that runs on it — model, pipeline and GPUs under one contract.
Tell us who will run the hardware, where and for what — we check how the project qualifies and come back with the platform, a bill of materials and a realistic timeline.
Not a quote? Ask a question