When I wrote the first part of this series, “From Land to Data Centre to GPU Financing: The Next Frontier for Indian Lenders”, GPU leasing at GIFT City, Gandhinagar, was still only a proposal. I asked whether it could offer lenders an alternative route to finance the chips inside a data centre.
The answer arrived on November 3, 2026, when the International Financial Services Centres Authority (IFSCA) Notification No. IFSCA/GN/2026/017 dated October 22, 2026, was published electronically in the Official Gazette; it was uploaded to the IFSCA website the next day. The operating lease of graphics processing units (GPUs) and “connected data centre equipment” is now a financial product in the International Financial Services Centre (IFSC). The route exists. The harder question is how lenders should price it.
It adds GPUs to the equipment covered by a Central Government notification of 14 December 2021 (S.O. 5199(E)), which treats the operating lease of IFSCA-specified equipment, including a hybrid of an operating and a financial lease, as a financial product.
The definition looks ahead. It covers GPUs, AI accelerators, tensor and neural processing units, “comparable computing or acceleration technologies” and rack-scale AI systems. Servers, storage, networking and power equipment qualify only insofar as they are dedicated to and directly associated with GPUs; equipment serving the data centre as a whole is excluded.
Part 1 of my article on GPU Financing identified a structural gap: the developer’s lender has no security over the GPUs, and the equipment financier has no rights over the building. Leasing narrows that gap, because the lessor owns the GPUs.
Lease terms can follow the technology cycle, a US dollar platform matches how GPUs are priced, and GIFT City’s aircraft and ship leasing framework offers a base to build on. For IFSC Banking Units, new business opens up: lessor facilities, sale-and-leaseback and advance payment financing.
But ownership does not stop the depreciation clock. A building financed today may stay useful for decades; a GPU may lose much of its commercial value in two or three years. Lender or lessor, the debt must be repaid faster than the asset ages. Leasing moves that risk. It does not remove it.
It is tempting to assume that GPU financing will simply follow the data centre: that the lenders who financed the buildings, and the documents they used, will extend naturally to the chips inside. The assumption does not hold. A data centre loan rests on land, a long-lived building and a mortgage that can be registered and enforced. A GPU loan rests on sophisticated movable equipment that may lose most of its value before the loan is half repaid, installed in a facility the borrower may not control, subject to export licences, and earning revenue from customers who may themselves be at an early stage.
As I noted in Part 1, a lender financing a data centre developer may have no security over the GPUs, and its existing precedents may not cover them. No lender enters this market with a ready-made template, nor does any previous mandate resolve the issues. Everyone at the table faces new questions.
Borrowing another lender’s term sheet will not do; no two GPU businesses carry the same risks. In Part 1, I said that ownership, access, contracted revenues, security and recovery must work together. That becomes a test every proposal, loan or lease, should pass before it reaches committee. A proposal that fails any part of it needs restructuring, not approval.
A familiar pattern
Indian lending has seen this pattern across successive cycles: the infrastructure boom around 2010-2015, the stressed-asset years, the insolvency cycle from the introduction of the Insolvency and Bankruptcy Code in 2016 to its peak in 2022, and the subsequent growth in funding for infrastructure, renewables and data centres. Each began with strong demand and apparently robust collateral. The real test came later, sometimes decades later, when contractual rights were tested against an evolving legal and regulatory landscape before courts.
Structure around the downside
The lenders who fared best were rarely the earliest entrants or those relying on familiar documents. They began by asking what would happen to their rights and security if the business failed, then structured the transaction around that answer, not the previous deal. What mattered was advice and involvement from experienced senior advisers in those deals, who could assess risks across the transaction, anticipate different scenarios and build safeguards into the structure, not merely familiarity with an existing set of documents or a standard template.
GPU financing is at the start of its own cycle. Demand is visible, from AI developers to the IndiaAI Mission. Discipline must arrive with it.
Behind the test lie questions of structure, tax, insolvency, regulation and enforcement, several of them left open by the notification. None is a reason to stay away. A committee does not need every question settled before it lends; it needs to know which risks can be structured, which can be priced, and which should not be accepted at all.
The lenders and lessors who work through these questions early will set the terms on which others follow. India’s AI ambitions depend on capital reaching computing infrastructure with confidence and multiple well-structured transactions in this space.
IFSCA has supplied the product and the venue. Clarity on domestic leasing and tax, and a legal framework that keeps pace, will turn it into a working asset class.
In Part 1 of my article, I set out one principle: security is only as good as the commercial arrangements supporting it. Part 2 adds its corollary: ownership is only as good as the time it buys.
Every new asset class in Indian finance has begun the same way: a few conversations about an asset no one in the room had financed before. GPU financing is at that stage now. It is among the most consequential conversations in Indian finance today, and one I look forward to continuing with lenders, lessors and others working through it. Those who price the clock before they say yes will shape the market.
About the author: Zeeshan Khan is a Partner and Head, Banking & Finance, at Fox Mandal & Associates.
This is the second part of his series on GPU financing in India.
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.