(Left) Anil Chalamalasetty, Group Chairman, AM Group, and Mahesh Kolli, Group President, AM Group.
AM Intelligence, an AI infrastructure company backed by Greenko’s promoters, will set up a 30-MW artificial intelligence compute facility in Hyderabad by March-April 2027 as the first step in its plan to build 1 GW of Compute-as-a-Service (CaaS) AI infrastructure.Under the CaaS model, foundation-model developers and enterprises can contract dedicated GPU clusters for periods ranging from one to five years for training and inference.The company plans to develop a total data centre capacity of 5 GW, of which 1 GW would support its own Compute-as-a-Service operations. The remaining capacity would be offered to hyperscalers, AI cloud companies and other operators.The 30MW deployment will be part of the first phase of the project, in which the company will build 200 MW of AI compute infrastructure with an investment of $8 billion over the next 18 months.“We have placed the order for 9,000 NVIDIA Vera Rubin GPUs for the first 30 MW in Hyderabad. The racks should start arriving in January, and we expect the facility to be fully operational by March or April 2027,” Mahesh Kolli of AM Intelligence told businessline.The Hyderabad facility will be followed by capacity additions in India, Malaysia, Finland and the US as the company scales up its AI compute operations to 200 MW over the next 18 months. The 200-MW rollout forms part of its longer-term plan to build 1 GW of Compute-as-a-Service capacity.“Of the proposed 5-GW data centre portfolio, we plan to retain about 1 GW for our own compute-as-a-service operations. The remaining capacity can be offered to hyperscalers, AI cloud companies and other global operators,” Kolli said.The larger 5-GW programme will span locations in India and overseas, with Noida, Visakhapatnam and Maharashtra among the locations being considered in India, as well as Finland and the US.Unlike conventional cloud companies offering storage, software hosting, and other enterprise services, AM Intelligence plans to focus on the high-density GPU infrastructure required to train and run artificial intelligence models.Asked about growing concerns over the power and water requirements of large data centres, Kolli said the newer AI infrastructure would have a substantially different cooling profile. The Vera Rubin-based systems would rely predominantly on liquid cooling, with about 95 per cent of the cooling handled this way, reducing the requirement for large volumes of water, he said.Published on August 28, 2026
