GPU and Data Centre Equipment Leasing in GIFT IFSC: IFSCA Proposes a New Financial Product

The International Financial Services Centres Authority (IFSCA) has proposed another significant expansion of the leasing ecosystem in GIFT IFSC by enabling the operating lease, including any hybrid of operating and financial lease, of GPU and connected data centre equipment as a financial product.

The proposal comes at a time when artificial intelligence infrastructure is rapidly becoming a major global capital-intensive asset class. High-performance GPUs, AI accelerators, servers, storage systems, networking equipment and data-centre infrastructure require substantial investment, while technology refresh cycles are becoming increasingly shorter.

Through its public consultation paper, IFSCA proposes to bring such equipment within the financial-product framework applicable in IFSC, potentially creating a new avenue for financing AI and data-centre infrastructure through GIFT IFSC. The proposal is presently at the consultation stage and will become operational only after the relevant notification and regulatory framework are finalised.

What Does IFSCA Propose?

The proposal derives from the Government of India notification S.O. 5199(E) dated December 14, 2021, under which operating lease, including any hybrid of operating and financial lease, of products or equipment specified by IFSCA may be treated as a financial product.

Such leasing activity is permitted under the IFSCA (Finance Company) Regulations, 2021. IFSCA has already used this regulatory route for leasing activities relating to aircraft, ships and, more recently, oilfield equipment.

IFSCA now proposes to add “GPU and connected data centre equipment” to this framework.

Importantly, the proposed equipment may be held by the financial institution in IFSC either:

• on an ownership basis; or
• through a lease-in, lease-out arrangement.

Following consideration of stakeholder comments, the proposed activity may be incorporated within IFSCA’s broader Framework for Leasing Activity in the International Financial Services Centre, with modifications considered necessary for this asset class.

What Is “GPU and Connected Data Centre Equipment”?

The proposal is wider than merely leasing individual graphics processing units.

IFSCA describes GPU and connected data centre equipment as accelerated computing devices together with the servers, storage, networking, interconnect and power equipment used directly or indirectly in their deployment and operation in a data-centre environment.

The draft notification proposes a technology-neutral approach covering, among other things:

• Graphics Processing Units (GPUs);
• General-Purpose Graphics Processing Units (GPGPUs);
• Artificial Intelligence accelerators;
• Tensor Processing Units (TPUs);
• Neural Processing Units (NPUs);
• GPU and AI servers;
• rack-scale AI systems;
• storage equipment;
• networking and interconnect equipment;
• server power supply units; and
• parts and accessories forming part of such systems.

The technology-neutral drafting is important because AI computing technology is evolving rapidly. The intention appears to be to accommodate successive generations of accelerated-computing systems without requiring a separate regulatory notification each time the underlying technology changes.

Why Is GPU Leasing Emerging as a Major Financing Opportunity?

IFSCA’s consultation paper places the proposal within the broader growth of the global AI infrastructure market.

According to the figures referred to by IFSCA, the global GPU market was approximately USD 65 billion in 2024 and is projected to reach approximately USD 400–500 billion by 2030.

The consultation paper further refers to an estimated USD 6.7 trillion of global data-centre capital expenditure by 2030, of which approximately USD 5.2 trillion may be attributable to AI-capable infrastructure. The market for accessing compute capacity rather than owning it is also expected to expand significantly.

The financing characteristics of GPUs are different from those of many conventional IT assets. High-performance AI equipment is expensive, supply can be constrained, and technology generations change quickly. Consequently, the principal economic risk may not necessarily be physical deterioration of the equipment, but technological obsolescence and residual-value erosion.

These characteristics make leasing commercially relevant because specialist lessors may be better positioned to finance the assets, assess residual values, manage refresh cycles and redeploy equipment across different users and markets.

Why Is the Opportunity Particularly Relevant for India?

India’s demand for AI computing infrastructure is expected to increase materially over the coming years.

IFSCA notes that India currently accounts for less than 5% of global AI-optimised compute power, while the United States and China together account for more than 70%. India’s AI GPU capacity is, however, expected to grow considerably.

The consultation paper refers to potential deployment of approximately 650,000–700,000 GPUs in Indian data centres over the next five years, representing an estimated USD 23 billion investment opportunity.

At present, India is also substantially import-dependent for this class of equipment.

While shared public compute infrastructure under the IndiaAI Mission is primarily directed towards startups, researchers, academia and public institutions, commercial demand from cloud service providers, colocation operators, hyperscalers and enterprises will have to be met through the market.

IFSCA therefore considers leasing an appropriate mechanism to bridge part of this financing requirement.

Why Leasing Instead of Direct Ownership?

The economics of AI infrastructure make leasing particularly relevant.

High-end GPUs and integrated AI systems require substantial upfront capital expenditure. At the same time, new generations of computing hardware can significantly improve performance and energy efficiency within relatively short periods.

An operating lease can therefore convert a large upfront investment into periodic lease payments and provide businesses greater flexibility in managing their technology infrastructure.

IFSCA specifically identifies several advantages of leasing, including:

• reducing the need for substantial upfront capital expenditure;
• transferring part of the obsolescence and residual-value risk to the lessor;
• enabling specialist lessors to manage and remarket equipment;
• facilitating redeployment of assets between lessees; and
• allowing users to respond to changing computing requirements.

The consultation paper observes that successive generations of GPU technology have delivered material improvements over periods as short as two to three years, making obsolescence risk particularly important in determining asset values.

Equipment Leasing vs GPU-as-a-Service: An Important Distinction

One of the most important aspects of the proposal is the distinction between leasing identified GPU equipment and providing GPU-as-a-Service (GPUaaS).

Under an equipment-leasing arrangement, a lessor owns identified GPU hardware and leases that equipment to a cloud operator, colocation provider or enterprise. The structure may also include sale-and-leaseback arrangements.

For the arrangement to constitute a lease under Ind AS 116, the lessee must have the right to control the use of an identified asset.

Accordingly, where specific GPU equipment is identified and the customer controls its use, the arrangement may constitute a lease.

GPUaaS operates differently. In such arrangements, the provider generally retains a pool of computing hardware and sells access to computing capacity. The underlying hardware may be substituted by the provider, and the customer is essentially purchasing computing capacity rather than controlling a specific identified asset.

IFSCA has expressly clarified that the current proposal is intended to cover identified equipment leasing and not GPUaaS service arrangements.

What Equipment Could Potentially Be Covered?

Annexure II of the consultation paper divides the proposed equipment into two broad categories.

Core GPU and Connected Data Centre Equipment

The proposed core category includes GPU and AI servers, compute nodes, storage units, rack-scale accelerated-computing systems, GPUs, TPUs, NPUs, AI accelerators, server power supplies, networking equipment, interconnect systems, accelerator boards, modules and related parts.

Allied Power and Thermal Equipment

IFSCA has also sought stakeholder views on including equipment that supports the operation of leased computing infrastructure, such as:

• high-capacity fans and air-movement equipment;
• precision data-centre air-conditioning systems;
• chillers and liquid-cooling equipment;
• immersion-cooling systems;
• power distribution units;
• busways; and
• dedicated switchboard panels.

This is an important area of the consultation because modern AI computing infrastructure is highly dependent on specialised cooling and power-management systems. IFSCA has specifically invited views on whether such equipment should be included and the parameters that should determine eligibility.

Conclusion: From Traditional Asset Leasing to AI Infrastructure

IFSCA’s proposal represents an important evolution of the GIFT IFSC leasing ecosystem.

After enabling leasing of aircraft, ships and oilfield equipment, the proposed inclusion of GPU and connected data centre equipment could extend IFSC leasing into one of the fastest-growing areas of global infrastructure investment.

The proposal is particularly significant because GPU financing is not simply conventional equipment financing. Rapid technological obsolescence, high capital costs, supply constraints, residual-value risk and the need for frequent technology refreshes create a strong commercial case for specialised leasing structures.

However, the framework remains a proposal under public consultation. The final scope, eligible equipment and regulatory conditions will depend upon the notification and leasing framework ultimately issued by IFSCA.