AMD, Cisco and HUMAIN have formed a strategic joint venture to develop up to 1 gigawatt of AI data center infrastructure in Saudi Arabia by 2030. The initial 100-megawatt phase is operating with AMD Instinct MI355X GPUs, AMD EPYC CPUs and Cisco networking, providing GPU-as-a-service for model training and inference to customers including Luma AI. The companies said the project was formally announced on Nov. 19, 2025, while a separate Aug. 31, 2026 update said the first segment had entered operations. The original project description identified AMD Instinct MI450 series GPUs for the initial phase, creating a hardware discrepancy with the later MI355X operational update. Up to 250 MW of additional capacity is planned from 2027, using next-generation AMD hardware, with the venture targeting as much as 1 GW by 2030. HUMAIN operates under the umbrella of Saudi Arabia’s Public Investment Fund and is pursuing a broader AI strategy that includes partnerships with Microsoft and Qualcomm. AMD and HUMAIN had previously signed a $10 billion agreement in May 2025 focused on expanding Saudi data center capabilities. The venture says its open platform is intended to support sovereign AI by giving governments, enterprises, research institutions and developers greater control over data location, model customization, deployment and governance, while emphasizing lower total cost of ownership than traditional AI infrastructure.