TL;DR
AM Intelligence has ordered 20,000 additional NVIDIA Rubin GPUs for roughly $4 billion, bringing its total committed frontier compute to 29,000 Rubin GPUs and nearly 100 MW, making it one of the largest non-North American buyers of next-gen NVIDIA silicon.
What happened
- 20,000 NVIDIA Vera Rubin NVL72 GPUs ordered across two new deals, priced at approximately $4 billion, for deployment in Malaysia and India.
- Delivery is scheduled for Q2 2027, adding 70 MW of follow-on capacity to AMI's existing Hyderabad commitment.
- The Hyderabad order, announced August 25, covered 9,000 Rubin GPUs at AMI's first AI factory, bringing the three-order total to ~29,000 GPUs and ~100 MW.
- All sites use NVIDIA Vera Rubin platform with RoCE networking, liquid cooling, and high rack-power density designed to absorb future GPU generations in the same footprint.
- AMI is backed by Greenko founders Anil Chalamalasetty and Mahesh Kolli, and will power compute with clean energy from AM Green and Greenko's 50 GW renewable infrastructure.
Why it matters
- AMI will operate the largest operational compute capacity in India by Q2 2027, ahead of domestic rivals Yotta, Adani, Reliance, Tata, and Larsen & Toubro.
- The $6 billion already committed for 100 MW is a fraction of AMI's stated roadmap: $20 billion more for an additional 300 MW coming online within 15 months.
- The full pipeline targets 1 GW of Compute-as-a-Service and 5 GW of powered AI data centres across India, the US, and Europe, with a Uttar Pradesh MoU alone targeting 500,000 high-performance chips by 2030.
- India's total installed data centre capacity is currently 1.2 GW and is projected to reach 10 GW by 2030, meaning AMI alone could represent a significant share of that buildout.
- Clean-energy-backed "Indian electrons for Indian tokens" positions AMI directly for India's sovereign AI initiatives and the government's IndiaAI Mission, which carries a 10,371 crore rupee budget over five years.
What to watch next
- Whether AMI secures anchor customers (hyperscalers, sovereign AI programs, frontier labs) before Q2 2027 delivery, validating the Compute-as-a-Service revenue model.
- Progress on the additional 300 MW tranche and whether the $20 billion capex commitment attracts co-investors or debt financing at scale.
- NVIDIA's broader India strategy: Rubin partnerships with Reliance, Tata, and Hero MotoCorp signal a competitive multi-partner race for India's GPU infrastructure crown.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.