AM Intelligence orders 20,000 more Rubin GPUs for India and Malaysia

AM Intelligence says it has placed two additional firm and binding orders for 20,000 NVIDIA Rubin graphics processing units, extending its planned AI computing infrastructure into India and Malaysia with deliveries scheduled for the second quarter of 2027.

The AI infrastructure platform, promoted by Greenko's founders, said the additional systems would provide approximately 70 MW of capacity. They follow an August commitment for 9,000 Rubin GPUs at its first Hyderabad AI factory, taking the total ordered to about 29,000 GPUs and close to 100 MW.

These figures describe ordered equipment and intended capacity, rather than commissioned computing services. The latest addition is part of AMI's approximately 400 MW Compute-as-a-Service pipeline, with further phases planned for 2027 and early 2028. The company also has longer-term ambitions for 1 GW of computing services and 5 GW of powered data centres.

AMI said the new facilities would follow the Hyderabad site's architecture, combining NVIDIA Vera Rubin systems with high-throughput networking, storage and liquid cooling. The intended customers include cloud providers, enterprises and AI developers.

NVIDIA's own product documentation describes Vera Rubin NVL72 as a rack-scale system combining 72 Rubin GPUs and 36 Vera CPUs with networking components. Deploying such systems therefore involves infrastructure beyond buying individual processors, including power delivery, cooling and networking.

AMI has not disclosed the India–Malaysia allocation of the additional GPUs, contracted customer volumes, site commissioning dates or a financing breakdown for the expansion.