From land to data centre to GPU financing: The next frontier for Indian lenders

India's infrastructure financing market has travelled a considerable distance, from roads, power plants and industrial projects to renewable energy and digital infrastructure. Data centres are the latest addition, driven by cloud computing and artificial intelligence (AI). While lenders are becoming familiar with financing land, buildings and associated infrastructure, the next challenge lies inside these facilities.

A data centre building may remain commercially useful for decades. The graphics processing units (GPUs) powering AI could become technologically outdated within a few years. Their value depends not merely on ownership, but on technological relevance, access to computing facilities and contracts generating predictable revenues.

GPU financing is emerging across Asia. A borrower may own neither the land nor the building. Instead, it acquires GPUs, houses them in a third-party data centre and earns revenue by providing computing capacity. The lender is effectively financing rapidly depreciating equipment located on premises the borrower does not control.

Indian lenders are increasing their exposure to data centre infrastructure, with institutions such as NaBFID and Tata Capital participating in sizeable financing transactions. Much of this activity, however, remains focused on land, construction and physical infrastructure.

Elsewhere in Asia, lenders are beginning to finance GPUs as a distinct asset class. For Indian banks, the question is whether conventional financing structures can accommodate equipment whose commercial life may be substantially shorter than the underlying loan tenure

Traditional data centre financing typically involves security over land, buildings, movable assets and receivables, supported by share pledges, escrow arrangements and sponsor undertakings.

GPU financing complicates this structure.

First, equipment may belong to AI companies or cloud providers occupying third-party facilities. The developer's lender has no security over those GPUs, while the equipment financier has no rights over the building.

Second, where an existing borrower raises separate GPU financing, lenders must address competing security interests, intercreditor arrangements and priority over customer receivables.

Even under a common financing structure, GPUs require separate valuation. A five-year loan against equipment that may lose much of its commercial value within two or three years creates an obvious mismatch. Accelerated repayment schedules, periodic valuations, mandatory prepayments and equipment replacement arrangements may therefore become important financing terms.

Enforcement presents further complications. Software licences may be non-transferable, customer data may attract statutory protection, and US export controls could restrict the resale of advanced chips.

A GPU lender is financing a computing business, not merely taking security over expensive equipment.

Repayment depends on customer demand, capacity utilisation, technology upgrades and the borrower's ability to generate sustainable revenues.

In India, demand is also supported by the IndiaAI Mission, which facilitates subsidised access to computing infrastructure through selected providers. Lenders must examine programme terms, payment arrangements and counterparty risks. Revenues from emerging AI businesses require particular scrutiny of their financial strength.

Loan documentation should therefore incorporate utilisation reporting, revenue coverage tests and mechanisms permitting equipment upgrades without compromising existing security.

1. Supplier agreement: Who owns the GPUs?

Suppliers may retain title until full payment. Lenders must establish when ownership transfers and whether warranties, buy-back arrangements and software licences remain available following enforcement.

2. Hosting agreement: Can the lender access its security?

GPUs housed in third-party facilities create operational dependencies. Hosting agreements should address access rights, termination notices, cure periods and the lender's ability to maintain operations or remove equipment following default.

3. Customer agreement: How secure is the revenue?

Long-term contracted payments offer greater debt-servicing certainty than arrangements terminable at short notice. Lenders must examine minimum payment commitments, service-level deductions, set-off rights, termination provisions and assignment restrictions. These contracts deserve scrutiny comparable to the financing documents themselves.

A dedicated project company holding the GPUs and principal commercial contracts may provide a cleaner financing structure.

Lenders should consider security over equipment and receivables, assignment of contractual rights, direct agreements with key counterparties, controlled collection accounts and debt-service reserves.

Direct agreements should provide notice of defaults, reasonable cure periods and appropriate step-in arrangements. A share pledge may offer an additional enforcement route, subject to applicable legal and contractual requirements.

Importantly, preserving the computing business may be more valuable than repossessing the equipment. Continued access to the facility, customer contracts and operational infrastructure could determine whether the lender recovers its exposure.

Security interests should be perfected and registered with the Registrar of Companies under the Companies Act, 2013 and with CERSAI under the SARFAESI Act, 2002 (SARFAESI), wherever applicable. Eligible secured creditors may invoke SARFAESI enforcement mechanisms, subject to statutory requirements.

The International Financial Services Centres Authority, regulator of India's international financial services hub at GIFT City, Gujarat, has proposed a framework for leasing GPUs and related data centre equipment. If adopted, leasing could provide an alternative to conventional secured lending, with lessors retaining equipment ownership. However, depreciation, contractual dependencies and recovery risks would remain. The proposal's final scope and commercial viability are yet to be determined

GPU financing represents another evolution in India's infrastructure lending market. Yet one principle remains unchanged: security is only as good as the commercial arrangements supporting it.

Expensive chips and reputed customers do not automatically make financing bankable. Ownership, access, contracted revenues, security and recovery must work together.

The challenge for lenders is applying established financing principles to an asset that could become commercially obsolete before the loan is repaid.

About the author: Zeeshan Khan is a Partner and Head, Banking & Finance, at Fox Mandal & Associates.

Disclaimer: The opinions expressed in this article are those of the author(s). The opinions presented do not necessarily reflect the views of Bar & Bench.

If you would like your Deals, Columns, Press Releases to be published on Bar & Bench, please fill in the form available here.