Every Gulf AI strategy runs on the same supply chain

The Gulf's AI programmes are described, consistently, as sovereign. Domestic models trained on domestic data in domestic data centres under domestic law. On the hardware and framework layers, every one of them runs on the same small set of international suppliers. Saudi Arabia's HUMAIN has local approval to buy 18,000 Nvidia accelerators, contingent on US government approval, and has a reported joint venture with AMD involving equity in a Saudi special purpose vehicle. Its targets run to 1.9GW by 2030 and six gigawatts by 2034. The UAE's Stargate project — a one-gigawatt cluster inside a planned five-gigawatt campus at Masdar City, built by Khazna Data Centers, a G42 company — lists OpenAI, Oracle, Nvidia, Cisco and SoftBank as participants. Its first 200MW phase is expected online in 2026. Abu Dhabi's TII builds Falcon and released Falcon-H1 Arabic in January 2026 on a hybrid Mamba-Transformer architecture. Saudi Arabia's SDAIA builds ALLaM with 400-plus experts and 160 government bodies. Qatar's QCRI builds Fanar. All train on Nvidia hardware, in frameworks written in the United States, on architectures published by international research labs. What sovereignty delivers in that arrangement is real, specific, and exactly what a regulator asks for. Data residency is genuine. A Saudi bank running ALLaM on HUMAIN infrastructure is not sending customer data to another jurisdiction, which is the point its supervisor examines. Model weights held domestically cannot be withdrawn by a vendor changing its terms. And a locally hosted model can be fine-tuned on data that could never leave the country. Those are the forms of control these programmes were built to secure, and they are the ones that determine whether a regulated institution can use AI at all. What the arrangement does not do is remove the supply relationship. Accelerator availability runs through an export licensing process, which is why HUMAIN's 18,000-chip purchase is described as contingent, and why the UAE structured Stargate with international companies as principals rather than as vendors. Giving a supplier equity in a project aligns its commercial interest with the licence being granted. That is the standard instrument for managing a supply dependency that cannot be removed, and both countries are using it. Removing it entirely would mean domestic fabrication or a separate supply chain. Leading-edge fabrication capacity exists in a handful of places, takes a decade to establish, and requires capital at a scale that is significant even for Gulf sovereign funds. The alternative supply chain carries its own access trade-offs, which is why no country in the region has pursued it. So the working description of Gulf sovereign AI is a strategy that takes control of the layers where control is available — data, weights, physical infrastructure, regulatory framework — and manages the remaining supply relationship through structure rather than through substitution. That is a coherent strategy and it is the same one every country outside the two chip-manufacturing blocs is running. The Gulf's version is distinguished by how much of the controllable stack it has actually built.