AM Intelligence (AMI) is aggressively scaling its high-density compute footprint across Asia to capture the burgeoning demand for sovereign and enterprise AI workloads. By securing two new binding orders for 20,000 NVIDIA Rubin GPUs, the Greenko-backed infrastructure platform is positioning itself as a major non-North American hub for frontier computing. These orders, which utilize NVIDIA Vera Rubin NVL72 rack-scale systems, expand AMI's total committed capacity to approximately 29,000 GPUs and nearly 100 MW. This expansion targets cloud providers and AI developers requiring massive scale in India and Malaysia.
AMI Secures 20,000 NVIDIA Rubin GPU Orders
The latest expansion adds approximately 70 MW of capacity to AMI’s deployment schedule, with deliveries for the new hardware slated for Q2 2027. These orders supplement an initial commitment of 9,000 NVIDIA Rubin GPUs announced on August 25, 2026, for the company's first AI Factory in Hyderabad. Consequently, AMI is establishing itself as one of the largest committed buyers of Rubin-class silicon outside of North America. To support this rapid scaling, the company is planning to bring an additional 300 MW to market over the next 15 months. This phase requires a projected capital expenditure exceeding $20 billion, building upon the $6 billion already committed for the initial 100 MW tranche. The company is leveraging its connection to Greenko’s renewable energy assets to manage the massive power requirements inherent in these large-scale deployments.
Liquid-Cooled Architecture and Global Pipeline
AMI is engineering its new facilities in India and Malaysia to match the architectural standards of its Hyderabad site. The infrastructure integrates the NVIDIA Vera Rubin platform with high-throughput RDMA over Converged Ethernet (RoCE) networking and advanced storage systems. To manage the intense thermal demands of next-generation silicon, the company is implementing liquid cooling throughout each site. This design choice aims to support high rack power density and allow the physical footprint to accommodate successive generations of AI hardware. Beyond the immediate 100 MW commitment, AMI is managing a global pipeline of approximately 400 MW of Compute-as-a-Service capacity across India, the United States, Europe, and Malaysia. The long-term roadmap targets 1 GW of Compute-as-a-Service and a broader program of 5 GW of powered AI data centers globally.
Key Takeaways
- AMI has secured binding orders for 20,000 NVIDIA Rubin GPUs to be deployed in NVIDIA Vera Rubin NVL72 rack-scale systems.
- The company's total committed capacity has reached approximately 29,000 GPUs and nearly 100 MW, with delivery scheduled for Q2 2027.
- AMI plans to invest over $20 billion in additional capacity over the next 15 months to expand its 400 MW global pipeline.
TechInsyte's Take
In our view, AMI is attempting to solve the most critical bottleneck in the AI era: the convergence of massive power availability and specialized silicon access. By vertically integrating Greenko’s renewable energy capabilities with NVIDIA’s next-generation Rubin architecture, AMI is not just building data centers; they are building "electron-to-token" pipelines. This strategy suggests that the next phase of AI infrastructure competition will be won by those who control the power source as tightly as the compute. The $20 billion capex target underscores the immense capital intensity required to compete in the sovereign AI and hyperscale markets, signaling a shift toward massive, energy-integrated infrastructure plays.
Questions & Answers
How does AMI plan to manage the extreme power density of NVIDIA Rubin systems?
AMI is implementing liquid cooling throughout its facilities and utilizing high-throughput RDMA over Converged Ethernet (RoCE) networking. This architecture is specifically engineered to handle high rack power density and ensure efficient data movement for large-scale AI workloads.
What is the scale of AMI's planned capital expenditure for future expansion?
Following an initial $6 billion commitment for the first 100 MW, AMI plans to invest more than $20 billion over the next 15 months to bring an additional 300 MW of capacity to market.
Which geographic regions are included in AMI's 400 MW global pipeline?
The company's current development pipeline for Compute-as-a-Service capacity includes locations in India, Malaysia, the United States, and Europe.
What is the strategic objective behind AMI's integrated energy and compute model?
AMI aims to deliver "electron-to-token" economics by integrating energy infrastructure with computing capacity. This model is intended to serve hyperscalers, frontier labs, and sovereign AI initiatives by providing scalable training and inference capabilities.
Source: AM Intelligence
