AMI Adds 20,000 NVIDIA Rubin GPU Orders, ~29,000 in Total

AM Intelligence (AMI), the AI infrastructure platform set up by the promoters of Greenko, said on 5 October 2026 that it has placed two further firm and binding orders for 20,000 NVIDIA Rubin GPUs. They are Vera Rubin NVL72 systems for Malaysia and India, with delivery scheduled for Q2 2027. AMI says that takes its committed Rubin capacity to approximately 29,000 GPUs and close to 100 MW.

Livemint, citing two people aware of the matter, puts the new orders at about $4 billion; AMI’s announcement gives no price for them.

What AMI Announced

AMI says the two further orders are for 20,000 NVIDIA Rubin GPUs, to be deployed as NVIDIA Vera Rubin NVL72 rack-scale systems. It says the expansion builds on its first AI factory in Hyderabad and is designed to give cloud providers, enterprises and AI developers access to large-scale computing capacity closer to where they build and deploy AI.

The orders are in addition to AMI’s inaugural commitment, announced on 25 August 2026, for 9,000 NVIDIA Rubin GPUs at the Hyderabad AI factory. AMI says the follow-on capacity totals about 70 MW, will be deployed in Malaysia and India, and has delivery scheduled for Q2 2027.

AMI says the three orders together bring its committed frontier compute to approximately 29,000 Rubin GPUs and close to 100 MW, “establishing AMI among the earliest and largest committed buyers of Rubin GPUs outside of North America.” That description is AMI’s own.

This publication’s own arithmetic: 9,000 plus 20,000 is 29,000, which matches AMI’s total.

What AMI Says Is Next

AMI says it plans to bring an additional 300 MW of capacity to market in the next 15 months, involving capital expenditure of “more than $20+ billion” in addition to the $6 billion already committed for the initial $100 MW.

It describes a pipeline of about 400 MW of Compute-as-a-Service capacity across India, the United States, Europe and Malaysia. It says the three orders placed to date are the first committed tranche, and that the balance is under development for delivery in 2027 and early 2028.

AMI also states an intention to bring 1 GW of Compute-as-a-Service capacity to global AI workloads, within a broader programme of 5 GW of powered AI data centres across India, the United States and Europe.

The Two Executives

Anil Chalamalasetty, AMI’s chairman, says in the announcement: “Both facilities will be built to the same architecture standard as the Hyderabad AI Factory, integrating the NVIDIA Vera Rubin platform with high-throughput RDMA over Converged Ethernet (RoCE) networking.” He adds that each site is “engineered for high rack power density and liquid cooled throughout.”

Vishal Dhupar, NVIDIA’s managing director for Asia South, says in the announcement: “AM Intelligence is bringing the next generation of AI factories to India and Malaysia with NVIDIA Vera Rubin.”

The Livemint Report

Livemint, citing two people aware of the development, says AMI placed a binding order for 20,000 NVIDIA Rubin GPUs in a deal valued at around $4 billion to set up 70 MW of capacity. It reports that an announcement would be made shortly. The $4 billion is Livemint’s sourcing, not AMI’s.

The Structural Read

The announcement uses four scales at once. GPUs (about 29,000), power (close to 100 MW committed, about 70 MW of it in the new orders), a development pipeline (about 400 MW, of which the three orders are the first committed tranche) and stated ambitions (1 GW of Compute-as-a-Service within a 5 GW programme). The GPUs and the committed power are described as orders; the rest of the pipeline is described as under development.

Money appears in the announcement only at the edges. The $6 billion is for the initial $100 MW and the $20+ billion is for the planned 300 MW; the announcement states no price for the two new orders. The only price this publication found is Livemint’s, about $4 billion, from two people it does not name.

AMI describes itself as “a strategic builder of long-term assets” that integrates energy infrastructure with the computing capacity required to train and run AI models, in pursuit of what it calls “electron-to-token economics.”

Three Implications

COMMITMENT VERSUS PIPELINE AMI’s own wording keeps ordered capacity apart from developing capacity: the three orders are the “first committed tranche” of the roughly 400 MW pipeline, and the balance is “under development.” The 300 MW addition and the 1 GW and 5 GW goals are stated plans and intentions, not orders.

POWER IS PART OF THE PITCH AMI’s announcement presents energy and compute as one business: the stated aim is integrating energy infrastructure with the computing capacity needed to train and run AI models, and Chalamalasetty says each site is “engineered for high rack power density and liquid cooled throughout.”

WHAT REMAINS UNKNOWN The announcement gives no price for the two new orders, no Malaysia and India split, no financing and no terms. This publication did not reach AMI’s own site and did not read NVIDIA’s side beyond its quote.

What Is Not Established

AMI’s announcement, as this publication read it, gives no price for the two new orders. The $6 billion refers to the initial $100 MW and the $20+ billion to the planned 300 MW. It does not give the split between Malaysia and India, the financing, or the terms of the orders.

This publication read the announcement as published by CXOToday and did not reach AMI’s own site. It did not read NVIDIA’s side beyond the quote, and it did not seek a response from AMI or NVIDIA.

The Bottom Line

AMI says it has placed two further firm and binding orders for 20,000 NVIDIA Rubin GPUs, about 70 MW with delivery scheduled for Q2 2027, taking its committed capacity to approximately 29,000 GPUs and close to 100 MW. The three orders are the first committed tranche of a roughly 400 MW pipeline whose balance AMI says is under development, and the 300 MW addition is a plan. The $4 billion price is Livemint’s, not AMI’s. This publication verified none of it independently.