India's AMI Locked In ~29,000 NVIDIA Rubin GPUs for AI Factories

So basically: On Oct. 5, Hyderabad-based AM Intelligence said it booked two more binding orders for about 20,000 NVIDIA Rubin GPUs. Add the earlier Hyderabad order and the company says it is now committed to roughly 29,000 Rubin GPUs and close to 100 MW of liquid-cooled AI factory capacity in India and Malaysia.

The orders

AMI’s first bite was Aug. 25, 2026: 9,000 Rubin GPUs for its first AI Factory in Hyderabad.

The Oct. 5 follow-ons add about 70 MW, set for delivery in Q2 2027, split between India and Malaysia. The systems are NVIDIA Vera Rubin NVL72 rack-scale builds.

Chairman Anil Chalamalasetty said the new sites will match Hyderabad’s architecture: Rubin compute, high-throughput RoCE networking for fast data movement, advanced storage, high rack power density, and liquid cooling throughout so later chip generations can reuse the same footprint.

NVIDIA’s Asia South managing director, Vishal Dhupar, said AMI is bringing next-generation AI factories to India and Malaysia by pairing accelerated computing with that networking and cooling stack.

AMI frames itself among the earliest and largest committed Rubin buyers outside North America. That is the company’s claim, not a third-party ranking.

The money and the pipeline

Even the firm 100 MW book is only the tip of AMI’s stated ambition. The company keeps pointing at a multi-country pipeline and Greenko-linked power. That combination is the real product pitch: not just GPUs in a hall, but GPUs next to electrons AMI’s group says it can control.

AMI says the first ~100 MW tranche already has about $6 billion committed. Over the next 15 months it wants another 300 MW online, with more than $20 billion of further capital spending on top of that $6 billion.

Beyond firm orders, AMI describes a ~400 MW Compute-as-a-Service pipeline across India, Malaysia, Europe, and the United States. That tranche is meant to feed a longer goal of 1 GW of CaaS inside a 5 GW powered AI data-center program spanning India, the U.S., and Europe.

The company is promoted by Greenko’s founders and leans on Greenko-owned renewable generation and storage. Marketing language centers on “electron-to-token” economics: tying power assets to training and inference capacity for hyperscalers, neo-cloud providers, sovereign AI projects, frontier labs, and local developers.

LiveMint separately described the new 20,000-GPU orders as a roughly $4 billion deal. AMI’s own Oct. 5 release does not state that dollar figure for the GPU purchase alone, so treat the $4 billion line as press reporting, not the primary source.

My take

This is a capacity booking story, not a rack that is humming today. Delivery is Q2 2027. The useful checks are whether the India and Malaysia halls energize on schedule, whether power and cooling match the liquid-cooled density claims, and whether 29,000 Rubin GPUs stay firm when every other region is chasing the same silicon and the same substations.