AM Intelligence orders 9,000 Nvidia Rubin GPUs for Hyderabad AI factory

AM Intelligence has placed a binding order for about 9,000 of Nvidia’s next-generation Rubin graphics processors for an artificial intelligence facility in Hyderabad, marking one of India’s largest announced purchases of advanced AI computing hardware.

The processors will be installed through Nvidia’s Vera Rubin NVL72 rack-scale systems and are scheduled for delivery in the first quarter of 2027, AM Intelligence said on Tuesday, 25 August.

The deployment will support the first 30 megawatts of computing capacity at the Hyderabad facility, which the company expects to become one of Asia’s earliest large-scale clusters based on Nvidia’s Rubin architecture.

AM Intelligence is the AI infrastructure business of AM Group, which was established by Greenko founders Anil Chalamalasetty and Mahesh Kolli. The group also includes renewable-energy company Greenko and green-fuels producer AM Green.

The Hyderabad project will form the first part of a planned one-gigawatt computing platform spanning India, the US, Finland and Malaysia. AM Intelligence plans to bring an initial 200 MW of capacity to market through capital expenditure of more than $8 billion, according to a company statement.

The company has separately outlined plans to develop five gigawatts of AI data-center infrastructure across India, the US and Europe. Its computing facilities will use liquid cooling to accommodate increasingly dense AI systems.

“For over two decades, we have focused on transforming electrons into value,” Chalamalasetty, chairman of AM Intelligence, said. “Today, the electron-to-token opportunity allows us to take this capability further, converting power infrastructure into frontier AI compute at scale.”

Each Vera Rubin NVL72 rack integrates 72 Rubin GPUs and 36 Vera central processing units. AM Intelligence’s order covers 9,000 Rubin GPUs deployed through those systems, rather than 9,000 complete NVL72 racks, as some initial reports suggested.

The Hyderabad cluster will use high-throughput networking and advanced storage systems to run large AI models, including models containing trillions of parameters. AM Intelligence estimates that the facility will provide about 450 exaflops of computing performance using Nvidia’s lower-precision NVFP4 format.

Nvidia says Vera Rubin can deliver 10 times the agentic-AI throughput of its preceding Grace Blackwell platform.

The chipmaker said in May that Rubin had entered full production, with suppliers including Dell Technologies, Hewlett Packard Enterprise, Lenovo and Supermicro building compatible systems at scale. Nvidia has not disclosed the value of AM Intelligence’s order.

The project extends the Greenko founders’ energy infrastructure strategy into computing, where access to large volumes of reliable electricity has become a key constraint on AI data-center development.