AM Intelligence (AMI) Orders 20,000 NVIDIA GPUs for India

AM Intelligence (AMI) orders the AI infrastructure company set up by Greenko’s promoters to announce two more firm and binding orders for 20,000 NVIDIA Rubin GPUs on October 5. They will run as Vera Rubin NVL72 rack-scale systems in India and Malaysia. Counting an earlier order for 9,000 GPUs, AMI has now committed to about 29,000 GPUs and close to 100 MW of capacity.

The numbers at a glance: AM

Detail	Information
Company	AM Intelligence (AMI)
New order	20,000 NVIDIA Rubin GPUs, placed as two binding orders
System	NVIDIA Vera Rubin NVL72
Locations	India and Malaysia
New capacity	About 70 MW
Delivery	Second quarter of 2027
Committed total	About 29,000 GPUs, close to 100 MW
Reported value	Around $4 billion, per unnamed sources

What did AM Intelligence (AMI) orders actually announce?

The new follow a first commitment on August 25 for 9,000 NVIDIA Rubin GPUs at AMI’s first AI factory in Hyderabad. The extra capacity totals around 70 MW, with delivery scheduled for the second quarter of 2027.

AMI says the aim is to let cloud providers, enterprises, and AI developers use large-scale compute closer to where they build and run their AI.

The announcement coverage I reviewed doesn’t say how many GPUs go to India and how many to Malaysia, or where exactly the sites are. Neither point has been confirmed.

Why does everyone want NVIDIA data center GPUs?

Developers train and run AI models on GPUs because GPUs can perform thousands of calculations at the same time. NVIDIA’s data center GPUs are the most common AI chips for that job, so a company’s GPU count and power capacity are a quick way to judge how much AI computing infrastructure it can offer. That is why an order like this one gets attention.

What is the Vera Rubin NVL72, and how is it different from Blackwell?

The Vera Rubin NVL72 is NVIDIA’s rack-scale system. The 72 stands for the number of GPUs linked inside a single rack so they behave like one big machine. It succeeds NVIDIA Blackwell, and reports describe it as far more advanced than Blackwell, with better power efficiency.

The two new sites will follow the same design as the Hyderabad AI factory, pairing the Vera Rubin platform with fast RoCE networking and advanced storage. RoCE moves data between GPU servers quickly, and that speed matters when thousands of chips share the work of running one model..

Who is behind AM Intelligence (AMI) orders?

AM Intelligence (AMI) orders comes from the promoters of Greenko, and it is based in Hyderabad. It plans to run its GPUs on clean energy and storage from AM Green and Greenko’s 50 GW platform. Cheap, steady renewable power is a big part of why the company thinks it can compete on AI data centers.

How much is the order worth?

Two people familiar with the deal told Mint’s sister outlet that the 20,000-GPU order is worth around $4 billion for 70 MW. AMI’s own release doesn’t give a price, so treat that figure as a report, not a confirmed number.

One of the same sources said AMI wants another 300 MW online in 15 months, which would cost about $20 billion. AMI has also proposed a $6 billion investment for 100 MW and has already deployed roughly $500 million of it. These numbers came from anonymous sources and may shift.

What else is AMI planning?

AMI says it has a pipeline of about 400 MW of compute-as-a-service capacity across India, the US, Europe, and Malaysia. Its stated goal is 1 GW of that service, as part of a wider 5 GW of powered AI data centers across India, the US, and Europe.

In January it also signed an MoU with InvestUP, the Uttar Pradesh government’s investment arm, for a 1 GW AI hub by 2030. The plan calls for $25 billion of investment and 500,000 high-performance chips.

What does this mean for NVIDIA GPUs in India?

One source summed up the plan as Indian electrons for Indian tokens, sold through GPU rental. If it works, Indian startups and companies could rent top-end AI GPUs at home instead of buying capacity abroad. That could help with data residency, latency, and cost.

None of it is available yet, though. The GPUs aren’t due until Q2 2027.

What could go wrong?

Delivering this on time depends on power, cooling, and networking all being ready, and that is hard at this scale. The financing figures are large and unconfirmed. AMI also needs enough customers to sign long compute contracts for the rental model to pay off. Until AMI names the sites, we can’t tell how much of the new capacity will serve India.

FAQs

How many NVIDIA GPUs has AM Intelligence ordered in total?
About 29,000. That is the new 20,000 plus 9,000 ordered in August.

When will the GPUs arrive?
The new orders are scheduled for delivery in the second quarter of 2027.

Where will they be installed?
In India and Malaysia. The exact locations haven’t been made public.

Are these NVIDIA Blackwell GPUs?
No. They are Rubin-generation chips. Blackwell is the previous generation.

How much power will the new orders add?
About 70 MW. With the Hyderabad order, committed capacity is close to 100 MW.

Who runs AM Intelligence?
Two people familiar with the deal put the 20,000-GPU order at around $4 billion for 70 MW. AMI’s own release doesn’t give a price, so treat that figure as a report, not a confirmed number. Its chairman is Anil Chalamalasetty.

Conclusion

AMI now has about 29,000 NVIDIA GPUs and nearly 100 MW of capacity on order across India and Malaysia, with delivery in Q2 2027. If it builds on schedule, India gets a large home-grown source of AI computing infrastructure. The next things to watch are the site locations and how the financing comes together.