The U.S.-UAE Agreement and the Gulf Compute Corridor

The Gulf states arrived at the AI contest with an unusual endowment: abundant cheap energy, sovereign wealth measured in the trillions, and a hunger to convert hydrocarbon rents into a post-oil economic identity. What they lacked was the one input money cannot manufacture on demand, namely access to the most advanced American accelerators, which sit behind United States export controls. The result has been a distinctive form of statecraft in which compute access is negotiated at the head-of-state level and bundled with security commitments. The United Arab Emirates has been the pace-setter.

The UAE's ambitions crystallised around Abu Dhabi's technology group G42 and a wider plan to build some of the largest AI data-centre capacity outside the United States and China. The obstacle was Washington's suspicion of the Gulf's deep commercial entanglement with China. American officials worried that advanced chips shipped to the UAE could be diverted, that Chinese firms embedded in Emirati networks could gain indirect access, and that sensitive model weights or hardware could leak eastward. The UAE's answer was to choose a side, at least in the domain that mattered. G42 publicly divested from Chinese holdings and stripped Chinese equipment from parts of its systems as a condition of deepening its American relationship, most visibly through a large investment from Microsoft.

What followed was a framework that treated chip access as a conditional, government-brokered privilege. The 2025 arrangements around a major Abu Dhabi compute build, sometimes described in terms of a five-gigawatt campus, were negotiated with explicit United States government involvement and security assurances. The template is worth stating plainly, because it is becoming the model for the whole developing world. A partner state gains access to advanced American compute. In exchange it accepts conditions: limits on Chinese technology and personnel, physical and cyber safeguards on the hardware, assurances against diversion, and often a commitment that a share of the value, the equipment, the cloud partnership or the capital flow, is anchored to American firms.

This is diplomacy conducted in units of processing power. It resembles nothing so much as the Cold War trade in security guarantees, where basing rights and weapons sales bound partners into a bloc. Substitute accelerators for fighter jets and the logic is identical. Access is the reward for alignment; exclusion is the penalty for hedging. The Gulf is the pilot region because it combines the money and energy to build at hyperscale with the geopolitical ambiguity that makes Washington insist on conditions. If the model works there, it becomes the standard offer to every capital that wants sovereign AI capacity and must choose whose chips to build it on.

The energy dimension is not incidental. The Gulf's comparative advantage is the ability to power enormous compute clusters cheaply, whether through gas, solar or, increasingly, planned nuclear. In a world where the binding constraint on AI is shifting from chips toward electricity, a region that can offer both the capital and the megawatts becomes structurally important. The United States, whose own grid struggles to add capacity fast enough for domestic demand, has an interest in allied compute being built somewhere with abundant power, provided that somewhere is securely inside its technological orbit rather than Beijing's.

Faraday's read is that the Gulf corridor is the purest expression of the course thesis. Export controls did not merely restrict; they created a new geography of AI infrastructure and a new instrument of alliance management. The controls forced the UAE to make a choice it might otherwise have deferred, and that choice pulled a strategically located, energy-rich, capital-rich state decisively toward the American compute bloc. The map was redrawn not by prohibition alone but by prohibition paired with a conditional offer of inclusion.

The risks are real and worth naming. Conditional access depends on trust, and trust in the Gulf is perennially tested by the region's continuing commercial ties to China and by the volatility of American policy across administrations. A partner that accepts today's conditions may resent them tomorrow, and a supplier that changes its terms with each election erodes the credibility that makes the bargain work. There is also a diversion risk that no contract fully eliminates; hardware is portable and know-how travels in people's heads. Still, the direction is set. For any state that wants to matter in AI, the operative question is no longer merely whether it can afford the compute, but whose bloc it is willing to join to be allowed to buy it. That is diplomacy, and compute is now its currency.