AM Intelligence Bets Big on AI Compute, Plans 1 GW Infrastructure

AM Intelligence is embarking on a significant expansion into AI infrastructure, with plans to establish a 30 MW artificial intelligence compute facility in Hyderabad by March-April 2027. This project forms the initial phase of a larger strategy to develop 1 GW of Compute-as-a-Service (CaaS) AI infrastructure globally.

The CaaS model is designed to cater to foundation-model developers and enterprises, offering dedicated GPU clusters for AI model training and inference on one- to five-year contracts. AM Intelligence envisions a total data center capacity of 5 GW, reserving 1 GW for its own CaaS operations and the remainder for hyperscalers and other cloud operators.

The first phase involves an investment of $8 billion over 18 months to build 200 MW of AI compute infrastructure. The Hyderabad facility alone will house 9,000 NVIDIA Vera Rubin GPUs, with racks expected to arrive in January 2027. Following Hyderabad, AM Intelligence plans capacity expansions in India, Malaysia, Finland, and the US.

Addressing environmental concerns, Mahesh Kolli of AM Intelligence highlighted that the new AI infrastructure will predominantly use liquid cooling for approximately 95% of its needs. This advanced cooling method significantly reduces the substantial water requirements associated with traditional data centers.