Morocco recently played Brazil to a draw at the 2026 World Cup. Close watchers of the game would not have been surprised had the North African giant won outright. Four years ago, when Morocco reached the semifinals in Qatar, the press received it as a fairy tale. What that coverage missed was the underlying implication: the geography of football competence had been quietly reorganizing for years before the established order noticed. The sport’s old cartography, in which European and South American dominance tracked so reliably with broader hierarchies of wealth and institutional depth that it barely seemed worth remarking on, has been redrawn.
What shifted was not aspiration alone but the patient accumulation of infrastructure, coaching depth, and institutional will in places the conventional rankings had written off. The same dynamic is now underway in artificial intelligence, and nowhere more consequentially than in the Middle East.
The standard account of who wins the AI race is structurally familiar: the United States and China have the chips, the compute, the talent, and the data. Everyone else consumes what they produce, on terms set in Palo Alto or Beijing. The Middle East plays the wealthy customer, purchasing frontier AI capabilities as an imported credential of modernity. The underlying assumption, shared across earlier technology debates, is that the established order is not seriously threatened from outside.
That assumption is wrong in a way the political economy of institutions can clarify. Daron Acemoglu and James Robinson have argued that those who control a system’s chokepoints resist relinquishing that control even where doing so would be efficient, because the chokepoint is what decides who else gets to compete at all. In frontier AI, the chokepoint is jurisdictional: whoever has legal authority over the company that owns a model has authority over the model itself, wherever in the world it runs.
Consider what happened this month: the US government imposed export controls on Anthropic, an American company, leading it to disable its two most capable models globally, citing the risk of access by foreign nationals. Beijing holds the equivalent authority over its own frontier labs, which must register their models with the state before public release. The chokepoint is a structural feature of the contest between Washington and Beijing, not a plot against any one country, but it is precisely the kind of concentrated control that capital independent of either jurisdiction is built to route around. The export controls on advanced chips were one signal. This is a clearer one.
SIGN UP FOR THIS WEEK IN THE MIDEAST NEWSLETTER
There is also a normative dimension the strategic debate tends to obscure. Large language models encode the assumptions, cognitive frameworks, and cultural logics of the data and institutions that built them. A model shaped by Silicon Valley’s priorities is not a neutral tool but a representation of the world from a particular vantage point. When a healthcare AI calibrated on American clinical data advises patients in Amman, it misrepresents the world to the people it is meant to serve. And as the Anthropic episode demonstrated, a model’s availability anywhere in the world depends on the security posture and political judgment of whichever single government holds jurisdiction over the company that built it. Diversifying the model ecosystem is therefore not only a competitive strategy but a normative goal: a question of whose cognitive frameworks, and whose jurisdictional authority, should govern the tools that increasingly mediate economic and social life.
Two structural facts make this plausible in the Gulf in a way earlier technology transitions did not. The first is capital independence. Gulf sovereign wealth funds, including the Abu Dhabi Investment Authority, Mubadala, and the Saudi Public Investment Fund, deployed more than $23 billion in AI and digital infrastructure in 2025 alone, with Mubadala’s $12.9 billion the largest single sovereign commitment to the sector globally. This is not portfolio diversification but the construction of financial capacity for AI that does not require any single government’s permission to deploy. It’s one answer, though not yet a complete one, to the jurisdictional chokepoint the Anthropic episode just exposed. Capital buys negotiating room, but the silicon and the frontier models that capital still has to purchase come from the same two jurisdictions the chokepoint runs through.
The second is the Arabic language gap. Arabic is spoken natively by more than 400 million people and is the fourth most used language on the internet, yet constitutes roughly 0.6 percent of training data in leading large language models. Every sector where language is the primary interface is underserved at enormous scale. Abu Dhabi’s Technology Innovation Institute has responded with Falcon Arabic, trained on native, non-translated Arabic text rather than the machine-translated data that dominates most multilingual models.
The strategic logic here is one Washington has consistently misread. The Gulf states’ simultaneous engagement with American AI companies and Chinese hardware suppliers looks like opportunism only if you expect binary alignment. When the technology-conscious Indian Prime Minister Jawaharlal Nehru built the Non-Aligned Movement during the Cold War, its critics called it equivocation. It was the opposite: an assertion of strategic agency, a refusal to subordinate national interests to either superpower’s terms. The Gulf states are constructing room for maneuver in a world bipolar at its center, where Washington and Beijing still set the frontier, and multipolar at its edges, where everyone else does the negotiating, with limited formal coordination among themselves to date.
The United Arab Emirates is building the infrastructure node for a Global South disinclined to inherit any single superpower’s technology stack. Saudi Arabia is embedding AI into the productivity logic of its Vision 2030 plan through the Saudi Data and Artificial Intelligence Authority, channeling oil giant Aramco’s decades of proprietary data into industrial coordination. Qatar is positioning around AI governance and diplomatic convening, an architecture no multilateral institution has yet credibly constructed. Three states converging on three different layers of the same opportunity by separate calculation, not design, is the more interesting development.
The talent pipeline remains the binding constraint, and the region knows it. Investments in institutions like MBZUAI in the United Arab Emirates and KAUST in Saudi Arabia echo, in spirit if not in design, the same wager Nehru made when he built India’s Institutes of Technology two generations ago, even if building deep research capacity remains a generational project.
None of this yet claims frontier capability; the labs in Palo Alto and Hangzhou hold that ground for now. What is taking shape is something narrower and, in the long run, more durable: states converting a multipolar moment into structural agency on their own terms, the same accumulation that let Morocco draw with Brazil, not by accident, but because the established order was watching the scoreboard and missed the academy system being built underneath it.
Bilal Baloch, PhD, is a nonresident senior fellow at the Atlantic Council’s MENA Futures Lab and a partner at Shorooq, an Abu Dhabi–headquartered venture capital fund. The views represented in this article are the author’s own.
Further reading
Image: A humanoid robot performs for attendees during an exhibition demonstration at Web Summit Qatar in Doha, Qatar, February 26, 2026. Matrix Images / Bojan Mustur
