The Middle East Is Quietly Building an AI Infrastructure Bloc

Hero photo: Dubai skyline at dusk — Robert Bock, Wikimedia Commons, CC0.

01 What is actually being built

Across the Persian Gulf, three of the world's richest states are pushing a build-out of artificial-intelligence infrastructure with no modern precedent outside the United States and China. Saudi Arabia, the UAE and Qatar collectively plan eight to ten gigawatts of AI compute across multiple sites — a bet that computational capacity will replace hydrocarbons as their main lever on the global economy.

The flagship projects are enormous. The UAE's plan — one of the biggest AI data center projects outside the U.S. — was originally envisaged as a 10-square-mile, 5-gigawatt campus in Abu Dhabi, with Nvidia, OpenAI and other U.S. firms as partners. Saudi Arabia's state-backed HUMAIN operates twenty computing centres and targets more than six gigawatts over the next decade. Qatar's investment authority is building positions of one to two gigawatts. This analysis is based on current reporting and records, not a prediction: the figures are announced targets from public reporting, and several have already been revised once — in the UAE's case, by war.

02 Three capitals, three strategies

Counterpoint's 2025 Global AI Cities Index ranked Dubai 4th in the world, Abu Dhabi 8th — and Riyadh 36th. LEAP 2026 was the week that distance stopped looking structural, because the three cities were never competing on the same thing.

Dubai won on adoption: an AI strategist embedded in every government department, AI training for every teacher, a programme to train a million engineers. Abu Dhabi won on research: MBZUAI opened in 2019 as the world's first university devoted entirely to AI, the Technology Innovation Institute wrote the Falcon models from scratch, G42 turned research into operating businesses, and 95% of the emirate's public sector is already AI-trained, with AED13 billion committed through 2027. Riyadh went after the layer it could scale furthest: energy, land and compute, with SDAIA holding strategy while HUMAIN moved silicon — 467 megawatts energized in Q1 2026 against a 1.5-gigawatt national target for 2030, and one week of LEAP commitments that included AMD systems already live, AWS's first Saudi cloud region due December 2026, a Microsoft region due November 2026, xAI opening at 50 megawatts scaling toward 500, and a 428-billion-parameter Arabic frontier model, humain-m3.

03 The war rewrote the engineering

The bloc's defining event was the war with Iran. Data centers became repeated targets in the U.S.–Iran conflict, and the Gulf's bet suddenly had to survive the region's own security realities.

According to Reuters, the UAE is quietly revising the 5-gigawatt Abu Dhabi project after the Iran war prompted a rethink of how and where critical infrastructure should be built: the original single 10-square-mile campus will likely become a network of data centers spread across the emirates, with authorities considering air defenses and underground construction to protect facilities from drone and missile attack. G42 says work is progressing as planned, with specifics under continuous review. Bloomberg reported that state-backed firms kept the project moving through the peak of the war, adding measures to protect workers.

The strategic lesson travels well beyond the Gulf: in a world where compute is critical infrastructure, concentration is a vulnerability. The most valuable AI campus outside the U.S. was redesigned precisely because it could be found on a map.

04 The American connection — and its strings

Every layer of the bloc runs through U.S. companies. Nvidia and OpenAI partner on the UAE campus; Microsoft's $15.2 billion UAE investment has been called the test case for U.S. AI diplomacy; AWS, Microsoft, xAI, AMD, Qualcomm and Cisco all took structural roles in Saudi Arabia in a single week. After the April ceasefire, the UAE and Saudi Arabia signed a combined $42 billion in sovereign AI compute agreements with three U.S. cloud providers — dedicated sovereign-cloud regions with on-soil data residency, locally operated key management, and preferential access to next-generation Nvidia and AMD accelerators.

Those strings run both directions. U.S. export controls still govern which chips reach the Gulf and in what order, which makes American policy an input to Gulf construction schedules. In exchange, Washington gets what it has wanted since the AI race turned geopolitical: allied compute capacity outside its own borders, financed by someone else's sovereign wealth. The Gulf gets the frontier. Whether that counts as sovereignty is the question the whole project hangs on.

05 Whose sovereignty is it, really?

Here the bloc's logic frays. Saudi Arabia commissioned humain-m3 from a Chinese lab and trained it further, rather than waiting for a domestic frontier model — buying a recipe is faster than writing one, but it leaves the Kingdom cooking from someone else's book. The deeper dependency is silicon: by Counterpoint's count, Nvidia's share of the chips behind the world's sovereign AI push is roughly 92%. The foundations of the Gulf build are national; the frontier is not.

Energy is the part that is genuinely theirs. Riyadh's digital economy is up 69% since the first LEAP, from $118 billion to $199 billion, and the Kingdom's comparative advantage — cheap domestic fuel, available land, patient sovereign capital — is exactly the 20-year layer of the AI stack that everyone else is short of. The Gulf is not buying its way into AI; it is selling the one input AI cannot do without.

06 Can three rivals form a bloc?

Calling it a bloc is the analytically risky move. The three states compete for the same partners, the same capital and the same title of preferred regional hub; officials on all sides describe coordination that has been promised more often than it has been practiced. Announced follow-ons — a joint UAE-Saudi training compute pool, a regional AI safety institute co-located between Abu Dhabi and Riyadh — would be the first real integration.

The constraints are physical. Eight to ten gigawatts of compute in an arid region means staggering demand for water and power, which is why NEOM, Masdar City and Kuwait's Green Mubarak Al-Kabeer City launched a joint platform on energy and water efficiency. And the demand case for the capacity is not primarily domestic — it is Europe, Africa and Asia, which makes the bloc a bet on everyone else's AI growth, not just the Gulf's.

07 The verdict

The verified facts: three Gulf states plan a combined eight to ten gigawatts of AI compute; the UAE's flagship 5-gigawatt campus is being redesigned as a distributed network after the Iran war; Saudi Arabia energized 467 megawatts in Q1 2026 against a 1.5-gigawatt 2030 target; roughly $42 billion in sovereign cloud agreements with U.S. providers were signed after the April ceasefire; MGX — Abu Dhabi's AI investment vehicle — closed its first fund at $49 billion in July.

The stakes: the Gulf is converting the last years of hydrocarbon leverage into the one input the AI economy cannot substitute — energy and land for compute — while renting the frontier from the United States. If the capacity gets built and stays standing, the region becomes the third pole of AI infrastructure. If the sovereignty layer stays foreign, it becomes the world's best-financed AI colony.