India and Malaysia are about to gain industrial-scale AI compute capacity.
AM Intelligence, the infrastructure company backed by Greenko's founders, has placed binding orders for about 20,000 of NVIDIA's Vera Rubin GPUs. That brings AMI's committed Rubin fleet to roughly 29,000 units, or nearly 100 megawatts of compute, and follows an earlier 9,000 GPU commitment for its Hyderabad AI factory. NVIDIA will deliver the new hardware as NVL72 rack-scale systems, and AMI says the additional roughly 70 megawatts are scheduled for delivery in Q2 2027.
This matters because these are physical machines being added where they will be used, not theoretical plans on a slide. Local businesses, startups and cloud providers can get onshore access to top-tier GPUs instead of routing heavy workloads overseas. That shortens delays, simplifies data residency, and lowers the practical barriers to building large AI services inside the region.
NVL72 rack-scale systems arrive as ready-made racks. Think of them as fully wired industrial shelves that include compute, networking and cooling. That design lets AMI pack more GPUs into each building and bring capacity online faster than assembling systems piece by piece.
For customers that need large-scale training and inference, AMI becomes a visible domestic supplier of Compute-as-a-Service. For the energy and real-estate teams in Hyderabad and Malaysia, this is a major industrial project that will demand serious power and cooling. It does not mean anyone can run these chips on a laptop; operating them requires data-center grade facilities and grid capacity.
The bigger question is execution: can AMI convert a roughly 400 megawatt pipeline into live services, and will local grids, supply chains and clients scale with it? Deliveries begin in Q2 2027, so the next year will show whether this is a one-off order or the start of a new regional AI infrastructure market.
