Ending a regulatory standoff that had frozen billions in infrastructure capital, the U.S. Commerce Department has authorized the export of 70,000 Nvidia GB300 chips to the UAE and Saudi Arabia.
The approval, granted under strict security protocols to prevent diversion to China, immediately operationalizes the region’s sovereign AI strategy, with Saudi champion Humain announcing today it has secured Adobe as its first global data center tenant.
This reversal allows G42 and Humain to finally deploy the massive compute clusters that were previously stalled by Washington’s national security concerns.
As previously covered by Winbuzzer, the deal follows months of intense negotiation, including a $15.2 billion investment plan that laid the financial groundwork for G42’s expansion.
The Regulatory Breakthrough: Chips for Sovereignty
Ending months of diplomatic limbo, the Commerce Department has issued export licenses for 70,000 Nvidia GB300 chips, split evenly between UAE’s G42 and Saudi Arabia’s Humain. Each entity is authorized to purchase 35,000 units, a volume sufficient to power multiple hyperscale data centers.
Specifically covering the Blackwell Ultra architecture, the authorization grants access to a significant upgrade over the H100s previously restricted. Such a move signals Washington’s willingness to share top-tier silicon with trusted allies, provided they adhere to stringent safeguards.
Approvals are contingent on strict compliance with a new “Intergovernmental Assurance Agreement,” which mandates rigorous end-use monitoring to prevent technology leakage to China. Representing a sharp departure from the Biden administration’s earlier stance, the policy shift reverses the “presumption of denial” that had characterized Middle East chip sales.
White House officials, including AI czar David Sacks, argued that withholding technology would merely push Gulf nations toward Chinese alternatives like Huawei, undermining long-term U.S. leverage. Under the deal structure, physical security audits of data centers ensure that U.S. personnel can verify the hardware remains in authorized hands.
Yousef Al Otaiba, the UAE’s ambassador to the U.S., framed the decision as a validation of the UAE’s strategic pivot away from Chinese infrastructure. He called the authorization a “milestone in the trusted and enduring partnership between our two nations.”
Operationalizing the Strategy: Humain’s Commercial Sprint
Immediately capitalizing on the regulatory green light, Humain has signed Adobe as its first global data center tenant, moving its $77 billion AI strategy from theory to revenue generation.
Under the strategic partnership announcement, Humain will host Adobe’s Firefly Foundry models on its sovereign cloud, addressing strict data residency requirements for enterprise clients in the Middle East.
By securing a major software vendor as an anchor tenant, Humain validates its ability to attract global workloads. Adobe will integrate Humain’s Arabic-centric “Allam” Large Language Model into the Creative Cloud ecosystem, enabling culturally nuanced content generation for regional users.
Simultaneously, Humain finalized a deal with Qualcomm to establish an AI Engineering Center in Riyadh, scheduled to open in December 2025. The engineering center plans involve deploying AI200 and AI250 rack solutions, with plans to roll out 200 megawatts of inference-focused capacity starting in 2026.
Adopting a multi-vendor approach allows Humain to hedge against Nvidia supply constraints while optimizing for different workloads. Training tasks will leverage the new Nvidia clusters, while inference operations can run efficiently on Qualcomm’s specialized hardware.
These agreements represent the first concrete operational steps for Humain since its formation by the Public Investment Fund (PIF) in May. Tareq Amin, CEO of Humain, emphasized the urgency of the market demand:
“The world is hungry for capacity. There are two paths you could take: you take it slow and we are definitely not taking it slow, or you go fast.”
The Regional Competition: G42 vs. Humain
G42 is leveraging the chip approval to accelerate “Stargate,” its 1GW compute cluster built in partnership with Microsoft and OpenAI. The UAE-based entity views the license as the key to unlocking its broader 5GW “UAE-US AI Campus,” which aims to serve as a digital embassy for global data.
Peng Xiao, CEO of G42, emphasized that the UAE’s infrastructure will mirror U.S. standards to ensure seamless interoperability for western tech giants. He committed to maintaining “symmetry and trust at every layer.”
Technically, the influx of GB300s provides a massive performance leap. Offering up to 1.5x faster inference than the H100 generation, the Blackwell Ultra architecture delivers a critical advantage for training the massive models envisioned by both nations.
Despite the Nvidia win, G42 continues to explore diversification. As reported by Winbuzzer, the company is holding active talks with AMD and Cerebras as part of a broader diversification strategy to prevent vendor lock-in.
Parallel build-outs in Abu Dhabi and Riyadh are creating a fierce regional competition for talent, power, and U.S. corporate tenants. Amin stated that Humain’s goal is to “redefine the silicon that powers the next era of generative AI.”
