India’s AM Intelligence Seeks $3 Billion Debt to Buy Nvidia Rubin AI Chips

AM Intelligence, an Indian artificial intelligence infrastructure company, is reportedly seeking to raise as much as $3 billion through bond sales and bank loans to finance purchases of Nvidia's next-generation Rubin graphics processing units, as competition to build AI computing capacity intensifies across Asia.

The Hyderabad-based company is considering a combination of debt instruments to fund the acquisition, Bloomberg reported on Friday. The final split between bonds and bank loans has not been determined, while AM Intelligence expects to receive approximately $300 million in advance payments from prospective customers that intend to secure computing capacity.

The company, which describes itself as one of the largest committed buyers of Nvidia Rubin GPUs outside North America, did not respond to a request for comment.

The proposed financing forms part of a much larger infrastructure expansion plan. AM Intelligence, owned by the founders of Indian renewable energy company Greenko, announced on October 5 that it would invest $6 billion to establish 100 megawatts of AI computing capacity across facilities in Malaysia and India.

The company plans to purchase approximately 29,000 Nvidia Rubin chips for the initial 100MW deployment. It also intends to bring an additional 300MW of capacity to the market over the following 15 months, requiring investment of more than $20 billion.

The scale of the planned expansion illustrates the growing capital requirements of the AI infrastructure business. Companies seeking to operate data centers equipped with advanced processors must secure substantial financing before generating revenue from customers renting computing capacity.

AM Intelligence's reported $3 billion fundraising highlights how AI infrastructure providers are turning to debt markets and bank financing to acquire expensive computing equipment.

The company's expectation of receiving $300 million in advance payments from prospective customers could provide an initial source of funding and indicate demand for its planned capacity. However, the proposed borrowing would still leave the company with significant financing obligations that must be supported by future revenue from computing services.

This business model is being embraced as enterprises and technology companies seek access to high-performance computing without purchasing and operating their own infrastructure. Providers such as AM Intelligence invest in GPUs, data centers, and supporting systems, then lease computing resources to customers through computing-as-a-service arrangements.

The model also creates financial exposure. Infrastructure providers must commit capital ahead of utilization, while their ability to service debt depends on securing customers, maintaining high equipment utilization and generating sufficient revenue from leased computing capacity.

The challenge is affecting companies investing in next-generation GPUs, where hardware acquisition represents only part of the total cost. Data centers also require reliable electricity, cooling systems, networking equipment, land, and operational infrastructure. Expansion plans therefore depend on more than access to Nvidia processors.

AM Intelligence says it is developing 5 gigawatts of AI data center capacity and 1GW of compute-as-a-service infrastructure across India, the United States and Europe. Within two years, it plans to establish 400MW of computing capacity.

The company's ambitions place it among a growing group of Indian businesses seeking to capture a share of the global AI infrastructure market. Their expansion is being supported by expectations that demand for computing power will continue rising as AI applications move from model development into commercial deployment.

The choice of Nvidia's Rubin architecture is key to AM Intelligence's investment strategy. Nvidia's next-generation GPUs are designed to support demanding AI workloads, including agentic systems that can execute tasks, coordinate software tools and operate with limited human intervention.

Neil Shah, vice president at market intelligence firm Counterpoint Research, said Nvidia's Rubin GPUs could give AM Intelligence an early advantage in India's AI computing market.

He said enterprises are expected to shift their attention towards running low-latency production agents next year, moving beyond the emphasis on foundational model training supported by Nvidia's Blackwell Ultra GPUs.

That transition could create demand for different configurations of computing infrastructure. Training large AI models requires substantial processing capacity, while running agents in production involves continuous inference, coordination, and interaction with software systems. Low latency and reliable access to computing resources become crucial when AI agents are expected to complete tasks in real time.

AM Intelligence is not alone in pursuing this market. Mumbai-based Yotta Data Services announced earlier this year that it planned to invest $2 billion in an AI hub equipped with Nvidia's Blackwell Ultra GPUs.

India's established energy and industrial groups are also moving into the data center business. Shah identified Adani Group, Reliance, and Tata Power among companies expanding into infrastructure that can provide computing capacity to hyperscalers and other customers.

The involvement of energy companies is remarkable because electricity availability is becoming a major constraint on data center expansion. Businesses with established energy assets or the ability to develop power infrastructure may have an advantage in securing the electricity required to operate large computing facilities.

AM Intelligence's expansion across India, Malaysia, the United States and Europe reflects the international nature of AI infrastructure investment. Companies are seeking locations that can offer access to customers, reliable power, connectivity and favorable operating conditions.

The expansion of domestic AI computing capacity is expected to help India reduce dependence on infrastructure located elsewhere while creating opportunities to serve international technology companies. The country's large technology services industry and growing enterprise market provide a potential customer base, although the economics of large-scale GPU deployments will depend on sustained demand.

The emergence of computing-as-a-service providers also creates a new competitive layer between chip manufacturers and the companies developing AI models and applications. Nvidia supplies the processors, infrastructure companies deploy and operate them, while hyperscalers and enterprises purchase computing capacity.

That arrangement allows infrastructure providers to participate in AI growth without developing their own foundational models. It also exposes them to the risk that demand forecasts may prove too optimistic, especially if customers delay deployments or newer hardware reduces the commercial value of existing equipment.

AM Intelligence's plan to raise billions in debt before completing its initial infrastructure rollout underpins the financial scale of the opportunity and the risks attached to it. However, its ability to convert advance customer commitments into long-term contracts and maintain high utilization rates will be important in determining whether its expansion can generate returns sufficient to support the borrowing.