Danny Weber
17:44 09-10-2025
© A. Krivonosov
The U.S. licensed Nvidia to ship 500,000 Blackwell GPUs to the UAE, anchoring an AI partnership tied to investments, data centers, and regional geopolitics.
The United States has officially granted Nvidia a license to deliver half a million of its latest Blackwell-series GPUs to the United Arab Emirates, the opening move in a sweeping technology partnership between the two nations. Bloomberg puts the value of the shipments in the tens of billions of dollars, but the accelerators will go not to local companies, rather to U.S. data-center operators running facilities in the Emirates.
The license builds on a May agreement allowing the UAE to purchase up to 500,000 Nvidia chips annually, including future Rubin and Feynman generations. In return, the Emirates pledged to invest $1.4 trillion in the U.S. economy over the next decade, on the condition that investments are matched on both sides. The initial phase excludes deliveries for the state-owned company G42, which is constructing a 5-GW data center for OpenAI; under the terms, however, G42 may later receive up to 20% of the processors designated for the region.
The decision signals a new strand of U.S. AI diplomacy, tying exports of critical hardware to partners’ political and investment commitments. Washington aims to deepen its influence in the Persian Gulf while curbing the spread of Chinese technology, notably Huawei, in a strategically vital region. The arrangement reads less like a straightforward hardware sale and more like a geopolitical pact.
Even so, the deal has drawn criticism. Opponents argue that the United States sharply raised the annual quota—from 100,000 to 500,000 GPUs—without adding extra oversight, while the UAE maintains close ties with China. The White House, for its part, treats the agreement as a template: if the program with the Emirates proves successful, similar accords could follow with other allies, helping Washington maintain a lead in allocating resources for the global AI infrastructure. The guardrails—or the lack of them—may ultimately determine how quickly that template travels.