AM Intelligence (AMI), the AI infrastructure platform established by the promoters of Greenko, has placed a firm, binding order for approximately 9,000 next-generation NVIDIA Rubin GPUs to be deployed at its premier AI factory in Hyderabad. Scheduled for delivery in the first quarter of 2027, the deployment utilizes NVIDIA’s Vera Rubin NVL72 rack-scale systems to power an initial 30 megawatts (MW) of capacity, positioning the Hyderabad facility as one of the very first frontier AI compute clusters in all of Asia.
Scaling Frontier Compute and Global Vision
The Hyderabad facility marks the first tranche of AMI’s ambitious Compute-as-a-Service platform, which spans a planned 1 gigawatt (GW) capacity globally across India, the United States, Finland, and Malaysia. As part of its near-term rollout, AMI plans to bring an initial 200 MW of capacity to market backed by an anticipated capital expenditure exceeding $8 billion.
Anil Chalamalasetty, Chairman of AM Intelligence, highlighted the strategic vision behind the initiative:
“For over two decades, we have focused on transforming electrons into value — addressing critical and emerging societal needs. Today, the electron-to-token opportunity allows us to take this capability further, converting power infrastructure into frontier AI compute at scale. Bringing latest generation NVIDIA Vera Rubin to India marks the next step in this journey and establishes AMI at the forefront of the emerging AI electron-to-token economy.”
Advanced Architecture and Performance Capabilities
Designed to run some of the world’s most sophisticated and massive models—including trillion-parameter systems and next-generation agentic AI applications—the Hyderabad AI Factory will integrate NVIDIA’s Vera Rubin architecture with high-throughput RDMA over Converged Ethernet (RoCE) networking and advanced storage.
Key architectural advancements include:
NVFP4 Precision Format: A new low-precision computing format designed to enhance workload efficiency.
Next-Generation HBM4: High Bandwidth Memory that drives performance upgrades.
Cost Efficiency: Engineered to reduce AI inference token-serving costs by up to 10 times compared to the preceding NVIDIA Grace Blackwell generation.
Massive Output: The Hyderabad cluster is designed to deliver approximately 450 exaFLOPS of NVFP4 inference compute capacity.
Building Toward a 5 Gigawatt Ecosystem
Looking beyond its initial milestones, AMI is developing a broader 5 GW portfolio of powered AI data centers across India, the United States, and Europe. These high-density facilities will utilize advanced liquid cooling to manage extreme thermal and power densities, ensuring compatibility with successive generations of AI silicon. By tightly coupling renewable energy infrastructure with cutting-edge NVIDIA compute, AMI aims to offer cost-competitive solutions for hyperscalers, neo-cloud providers, sovereign AI initiatives, and local developer communities.
