Nvidia reveals expansion of its cloud partners in Europe and Africa, with UAE-based G42 at the core of sovereign infrastructure

Nvidia announced that its network of certified cloud partners is continuing to expand across Europe, the Nordic countries, Africa and Central Asia, as a sign of an accelerating race for digital-sovereignty infrastructure. The expansion is not limited to adding compute capacity; it redraws the map of control over where data are processed and models are trained, and places entities such as the UAE-based G42 in the role of a strategic partner whose reach extends from Abu Dhabi to the heart of the European continent.

The details published on Nvidia’s official blog reveal three parallel pillars: sovereign infrastructure, open models, and secure agent boundaries. In the first pillar, Nscale, Crusoe, Aleria, Cassava Technologies, Firebird, G42, Radiant, SWI Compute and Vultr top the list of partners building regional AI factories on Nvidia’s full-stack DSX platform, including the Vera Rubin and Blackwell Ultra systems and the Spectrum-X networks.

Sovereignty Starts with Data Location DecisionThe link between these entities is not only technology but the ability to offer a genuine alternative to the US-centric cloud. When G42 chooses to deploy Nvidia-backed infrastructure in multiple European markets, it exercises a sovereign option: European and African enterprises’ data remain within a closely related jurisdiction, and the owning party controls access and security terms. The same pattern recurs with Cassava, which extends its network from Africa, and with Vultr, which is expanding its points of presence in Central Asia. The result is a distributed, decentralized network in which each partner decides where sensitive workloads settle.

In the models pillar, Nebius stands out for adopting the Groq 3 LPX platform for serverless and dedicated inference, while Mistral trained its Mistral Large 4 model on 3,800 Grace Blackwell units in European data centers it operates itself. Verda, Baseten and Fireworks, together with Era 4, complete the picture: a lightweight, customizable open-ecosystem running on infrastructure owned by regional operators.

Agent Boundaries: From Silicon to SoftwareThe third pillar, OpenShell, tackles the dilemma that follows deployment: how to ensure that an intelligent agent does not exceed its permissions? The reference architecture integrates Vera Rubin NVL72 for inference, Vera CPU with OpenShell for the toolchain, and BlueField-4 with Sentry technology for silicon-level execution. G42 is explicitly listed among the partners supporting this architecture, alongside CoreWeave, Crusoe, DigitalOcean, Firmus, Fireworks, Lambda, Lightning AI, Nebius, Nscale, Verda and Vultr.

What this means for readers in the Gulf, Egypt and the Levant: sovereign infrastructure is no longer an isolated national project but a cross-border network shared by regional operators bound by common technical standards. G42, which leads the Emirati sovereign axis, has become a node in a Europe-Africa network, giving institutions in the region geographically and jurisdictionally proximate deployment options and reducing reliance on data routes that pass through unaccounted spheres of influence. The next race will be not for the largest model but for who controls the platform on which decision-makers decide where their models run and under which laws their data are protected.