AI “factories” are spreading beyond North American hyperscalers. On 5 October 2026 Hyderabad-based AM Intelligence (AMI)—the AI infrastructure platform set up by promoters of Greenko—said it had placed two further firm and binding orders for about 20,000 Nvidia Rubin GPUs as Vera Rubin NVL72 rack-scale systems. [1]
All figures below are company-reported from AMI’s press release unless noted.
What did AMI order this time?
The new orders add roughly 70 MW of capacity to be deployed in Malaysia and India, with delivery scheduled for the second quarter of 2027. They sit on top of AMI’s 25 August 2026 inaugural commitment for 9,000 Nvidia Rubin GPUs at its first AI factory in Hyderabad. [1]
Together, AMI says the three orders bring committed frontier compute to approximately 29,000 Rubin GPUs and close to 100 MW—positioning AMI, in its own words, among the earliest and largest committed Rubin buyers outside North America. [1]
Capex and the next 300 MW
AMI plans to bring an additional 300 MW of capacity to market over the next 15 months, involving capital expenditure of more than $20 billion on top of $6 billion already committed for the initial 100 MW. [1]
Chairman Anil Chalamalasetty said both new facilities will match the Hyderabad AI factory architecture: Nvidia Vera Rubin with high-throughput RDMA over Converged Ethernet (RoCE) networking, advanced storage, high rack power density and liquid cooling throughout so successive silicon generations fit the same footprint. [1]
Nvidia’s Vishal Dhupar, managing director for Asia South, is quoted welcoming AMI’s Vera Rubin buildout for developers and enterprises in India and Malaysia. [1]
Pipeline claims beyond the orders
AMI describes a pipeline of about 400 MW of Compute-as-a-Service capacity across India, the United States, Europe and Malaysia. The three orders to date are the first committed tranche; the balance is under development for 2027 and early 2028 delivery, the company says. [1]
Longer term, AMI states an intention to bring 1 GW of Compute-as-a-Service to global AI workloads within a broader programme of 5 GW of powered AI data centres across India, the United States and Europe. The About section repeats Greenko-linked energy ambitions (including a 50 GW energy platform and large storage and green-ammonia targets); those are corporate roadmap claims, not part of the Rubin order itself. [1]
What we don’t know
- Company press only. Orders, megawatts, dollar capex and delivery dates are AMI’s statements; no independent confirmation of deposit, supply allocation or site permits appears in the release.
- Customer offtake. Which cloud providers or enterprises will rent the capacity is not named.
- Financing. How the “more than $20 billion” incremental spend will be funded is not detailed.
- Execution risk. Q2 2027 delivery and a further 300 MW in 15 months depend on Nvidia supply, power and construction—none of which the release audits.
The Bottom Line
AMI says two new binding orders for ~20,000 Nvidia Rubin GPUs (~70 MW in India and Malaysia, Q2 2027) lift committed capacity to ~29,000 GPUs and nearly 100 MW, with another 300 MW and more than $20 billion of further capex planned on top of $6 billion already committed. Treat every megawatt and dollar as company-reported until sites light up and customers sign.
Sources
- AM Intelligence press release, Hyderabad, 5 October 2026 — “AM Intelligence Expands NVIDIA Vera Rubin AI Factory Buildout in India and Malaysia”: https://amintelligencegroup.com/press-release-vera-rubin-expansion.html
