AM Intelligence orders 20,000 NVIDIA Rubin GPUs

In a press release dated October 5, 2026, AMI announced two further firm and binding orders for 20,000 NVIDIA Rubin GPUs.[1] The company said the GPUs will be deployed as NVIDIA Vera Rubin NVL72 rack-scale systems, with delivery of the new capacity scheduled for the second quarter of 2027.[1] Separately, Mint reported that the 20,000-GPU order was worth around $4 billion, citing two people familiar with the development. That figure is a media report, not an amount stated in the AMI facts provided here.[2]

The new orders join an order AMI announced on August 25, 2026, for 9,000 NVIDIA Rubin GPUs at its first AI Factory in Hyderabad.[1] AMI says all three orders together commit it to approximately 29,000 Rubin GPUs and close to 100 MW of capacity.[1] The company described the three orders as the first committed tranche of its pipeline; it said remaining capacity is under development for later phases in 2027 and early 2028.[1] Those figures describe AMI’s commitments and plans, rather than capacity already delivered.[1]

AMI says it is integrating the NVIDIA Vera Rubin platform with high-throughput RoCE networking and advanced storage systems.[1] It describes each site as engineered for high rack power density and liquid-cooled throughout, and says the infrastructure is designed to accommodate successive generations of AI silicon within the same physical footprint.[1] AMI says the intended customer base includes hyperscalers, neocloud providers, sovereign AI initiatives, frontier labs, AI-native companies and local developer communities.[1] The company says the expansion is intended to bring large-scale compute closer to where cloud providers, enterprises and AI developers build and deploy AI.[1]

AMI says it plans to bring another 300 MW of capacity to market in the next 15 months, involving more than $20 billion in capital expenditure in addition to $6 billion already committed for its initial 100 MW programme.[1] The company also says it intends to provide 1 GW of Compute-as-a-Service capacity for global AI workloads and that its wider programme includes 5 GW of powered AI data centres across India, the United States and Europe.[1] AMI describes itself as an AI infrastructure platform set up by Greenko’s promoters, with a model that integrates energy infrastructure and computing capacity for AI training and inference.[1]