AM Intelligence’s Hyderabad facility will bring online rack-scale Nvidia Vera Rubin systems next year to run trillion-parameter AI models and next‑gen agentic AI applications, powered by Greenko-backed energy infrastructure.

The Hyderabad clusters will deploy liquid cooling and high‑throughput networking, targeting roughly 450 exaFLOPS of NVFP4 inference compute for large-scale models.

NVIDIA Vera Rubin architecture and NVL4 advances, along with storage and memory improvements, are expected to enable efficient AI inference at scale, with AMI’s Hyderabad AI Factory aiming for about 450 exaFLOPS of NVFP4 compute for trillion‑parameter systems.

Founders’ energy‑infrastructure focus expands into computing, addressing electricity access as a constraint on AI data‑center development.

AM Intelligence is part of Greenko and is pursuing an $8 billion plan to create 1 gigawatt of computing capacity, with initial capacity already contracted to a US customer under NDA.

This order is part of the broader $8‑billion push to establish 1 gigawatt of compute capacity, leveraging Greenko’s renewable power to secure early demand.

The initiative positions AM Intelligence to deliver scalable AI compute at low cost through renewable Power, as part of a strategic 1 GW rollout.

Funding for the buildout will combine debt and equity, with a short capex‑to‑cash‑flow cycle and potential bond-market tapping.

AM Intelligence intends to finance the project via a mix of debt and equity, leveraging its balance sheet and experience in bond markets.

High‑end Nvidia compute clusters remain in tight supply due to persistent demand from data centers and major tech firms for AI training.

Nvidia Rubin systems are in strong demand and are contributing to multiyear backlogs as data centers and hyperscalers race to deploy next‑gen AI training capacity.

The plan envisions an initial 200 MW of AI compute capacity, with more than $8 billion in capex guiding a broader race to support ever larger AI models.