AM Intelligence Plans 30 MW AI Infrastructure Facility in Hyderabad by April 2027

AM Intelligence, an artificial intelligence infrastructure company backed by Greenko’s promoters, has announced plans to establish a 30-megawatt (MW) AI compute facility in Hyderabad. The project is scheduled to become fully operational by March or April 2027 and marks the first phase of the company's broader objective to build 1 gigawatt (GW) of Compute-as-a-Service (CaaS) AI infrastructure.

Under the CaaS operational model, foundation-model developers and enterprise clients can contract dedicated GPU clusters for durations ranging from one to five years to handle both training and inference workloads.

The initial 30 MW deployment forms part of a larger 200 MW AI compute infrastructure rollout. Mahesh Kolli, Group President at AM Group, stated that the company has placed an order for 9,000 NVIDIA Vera Rubin GPUs for the Hyderabad site, with the equipment racks expected to begin arriving in January.

Following the Hyderabad project, AM Intelligence plans to expand its capacity additions to other locations, including Malaysia, Finland, and the United States, to scale its operations to 200 MW over the next 18 months. This rollout is supported by an investment of $8 billion.

The company's long-term roadmap includes developing a total data centre portfolio of 5 GW. Out of this total capacity, AM Intelligence intends to retain 1 GW specifically for its CaaS operations. The remaining 4 GW of capacity will be made available to hyperscalers, AI cloud service providers, and other global operators.

Within India, the potential locations under consideration for the larger 5 GW programme include Noida, Visakhapatnam, and Maharashtra, alongside international sites in Finland and the US.

Unlike conventional cloud computing providers that primarily offer storage, software hosting, and general enterprise services, AM Intelligence is focusing specifically on high-density GPU infrastructure required for training and executing advanced artificial intelligence models.

Addressing industry concerns regarding the high power and water consumption typically associated with large data centres, Kolli noted that the newer infrastructure will feature a different cooling profile. The NVIDIA Vera Rubin-based systems will rely predominantly on liquid cooling, handling approximately 95 percent of the cooling requirements and significantly reducing the need for large volumes of water.

"The expansion of specialized AI infrastructure like the Compute-as-a-Service model addresses a critical requirement for enterprises and developers scaling heavy workloads. By focusing on high-density GPU deployment and advanced liquid cooling, companies can manage resource constraints more efficiently while meeting the heavy computational demands of foundational AI models." — Dr. Shishir Gupta, Founder & CEO, StartupLanes