The new supercluster builds on du's deployment last year of what was billed at the time as the largest Nvidia GPU supercluster at 5+MW capacity. With the latest expansion, the supercluster at du's DSO data center now sports 13+MW of operational capacity powered by Nvidia B300 GPUs.
Du said the expanded GPU cluster allows for higher computational density and improved thermal management, which means better performance capabilities for du's enterprise customers that plan to implement AI workloads, research initiatives, and commercial applications that require heavy compute resources.
Du also said the cluster uses advanced direct-to-chip liquid cooling technology, which significantly enhances energy efficiency while enabling optimal performance of the Nvidia B300 architecture.
“Deploying the most advanced GPU cluster technology available today will help us meet current market demands as well as anticipate and shape the computational requirements of tomorrow's AI-driven economy,” said du CEO Fahad Al Hassawi in a statement. “Our customers now have access to processing power that will accelerate their innovation timelines and competitive advantages.”
Mohamed Taha, executive VP at NextGenAI, added that the expanded facility represents a new era of AI accessibility in the Middle East.
“The combination of du's world-class infrastructure and our AI expertise creates an ecosystem where businesses can rapidly scale their AI initiatives without the traditional barriers of infrastructure limitations or deployment delays,” he said.
Du and NextGenAI unveiled the expanded AI supercluster at last week’s GITEX Global 2025 event in Dubai, at which AI was a prominent theme. Speaking onstage at the event last Monday, UAE Minister of Economy and Tourism Abdulla Bin Touq said AI must become a core pillar of government spending, on par with defence and cybersecurity, for nations to truly embrace the digital age.
During the event, du also launched two new sovereign AI offerings – an AI Park in Dubai that will feature several liquid-cooled hyperscale data centres delivering up to 1GW of capacity by 2030, and a National Hybrid AI platform that offers government entities and enterprises an integrated set of services to develop AI applications in their cloud of choice.
