Saudi Arabia Bets on Both AI Blocs, Turning Compute Into Leverage
RIYADH — Saudi Arabia is assembling one of the world’s largest AI buildouts, and it is doing so with American technology at its core. Amazon Web Services has committed $5.3 billion to data centers in the kingdom. Yet at this month’s LEAP conference in Riyadh, HUMAIN, the state-backed AI company, introduced Arabic-language technology built on a model from China’s MiniMax.
That may look contradictory. In Riyadh, it is a strategy.
The standard story of the AI race divides the world into two camps: the United States, home to Nvidia, the major hyperscalers and most frontier labs; and China, which has developed its own models, engineering talent and an increasingly competitive technology stack. Saudi Arabia and the UAE are not trying to recreate either ecosystem from the ground up. They may not need to. Their advantage lies in being one of the few places where both can operate at once.
The Gulf’s visible strengths are capital, energy and land. Its less visible one is geopolitical flexibility. Saudi Arabia is demonstrating how that works. HUMAIN is building infrastructure with American technology and has agreed that Chinese frontier models will not be trained on computing capacity it secures through US export approvals. At the same time, it is keeping its relationships with Chinese companies open. Beyond the MiniMax deal, ByteDance’s cloud unit is building infrastructure in Riyadh, Tencent Cloud has obtained a Saudi operating license, and Lenovo has already produced its first “Made in Saudi Arabia” laptop on a Riyadh assembly line.
The logic is straightforward. US restrictions push Riyadh toward Chinese alternatives; Chinese gaps in frontier performance pull it back toward Washington. Either way, both powers end up competing for something the Gulf increasingly controls.
For Washington, the risk is that export controls meant to keep advanced chips out of Chinese hands may also push Gulf partners to build parallel relationships with Chinese suppliers. For Beijing, the opportunity is that open models can win users even where its chips cannot follow. For the Gulf, the payoff is leverage: it can attract American capital and Chinese software while positioning itself as a neutral node in a fragmenting technology order.
AI can feel like software, but much of it behaves like heavy industry. It requires vast data centers, electricity, cooling, land and capital. Saudi Arabia has all five. DataVolt is already building a roughly $1 billion, 100-megawatt facility at NEOM designed to export compute, with plans to expand to 360 megawatts.
That points to a different model of sovereign AI than the one usually discussed. Saudi Arabia does not need to design the best chip or train the best model. It can own a growing share of the infrastructure on which other countries’ models run—and use access to that infrastructure to negotiate investment, localization commitments and technology transfer.
The stakes vary by capital. Washington now treats chip policy as foreign policy: access to Nvidia’s most advanced processors is less a commercial transaction than a form of access to sensitive military technology. Beijing views open models and a stronger domestic stack as a way into markets where its hardware cannot follow. Riyadh’s interest is simpler: not having to choose between them.
If that approach holds, the AI competition may resemble the Cold War more than expected. Instead of two sealed blocs, it could produce a handful of swing states able to play both sides, extracting economic and strategic benefits as Washington and Beijing vie for their allegiance.
The Gulf spent decades turning control over energy into geopolitical leverage. It is now applying the same playbook to compute.
Chinese chipmakers are starting to challenge Nvidia’s dominance. Shanghai Enflame Technology’s IPO was 6,000 times oversubscribed, Nikkei reported.