2 hrs ago
India Needs Specialized Financing for GPU-Powered Data Centres
India wants cloud companies to run businesses from Indian data centres.
It is offering them a long tax break on profits earned from customers outside India.
But data centres need expensive computer chips called GPUs, not just buildings and electricity.
GPUs can become outdated much faster than buildings do.
The article says buildings should be financed like long-lasting infrastructure.
Chips should instead be leased or financed through special loans and contracts.
India is considering allowing GPU leasing as a financial product in GIFT City.
The article also warns that unused GPUs could become like unused internet cables from an earlier investment boom.
India’s data centre capacity grew from about 375 MW in 2020 to around 1,500 MW in 2025.
The FY27 Budget offers foreign companies a tax holiday until 2047 for serving global cloud demand from India.
GPUs, servers, networking and liquid cooling can represent 60–80% of an AI data centre’s total cost.
REITs and InvITs can finance long-lived assets such as land, buildings, power and cooling systems.
The article proposes GPU leasing, vendor finance or loans backed by contracts, while warning against speculative overbuilding.
- Who
- The Indian government, cloud companies, data centre operators, infrastructure investors and financial regulators are involved.
- What
- India is encouraging global cloud businesses to operate from Indian data centres while considering specialized financing for GPUs and related equipment.
- Where
- India, including GIFT City and its International Financial Services Centre.
- When
- India’s data centre capacity figures cover 2020 to 2025; the FY27 Budget offers a tax holiday until 2047.
- Why
- The proposed tax incentives support future profits but do not solve the high upfront cost, rapid obsolescence and taxation issues associated with GPUs.
Specialized financing advocates
Overbuilding and financing-risk concerns
How to finance data centres
Specialized financing advocates
Separate long-lived buildings and infrastructure from short-lived GPUs, using REITs or InvITs for the former and leasing, vendor finance or contract-backed loans for the latter.
Overbuilding and financing-risk concerns
Financing GPUs against customer contracts can become dangerous if those contracts disappear or fail to generate expected usage.
Role of tax incentives
Specialized financing advocates
A long tax holiday can attract global cloud operations, but it should be paired with financing structures that address the cost of chips.
Overbuilding and financing-risk concerns
Tax incentives alone may encourage announcements without ensuring that the underlying assets are economically viable or fully used.
Key facts
- Data centre capacity
- About 375 MW in 2020, rising to around 1,500 MW in 2025.
- Tax holiday
- The FY27 Budget offers a tax holiday until 2047 for qualifying foreign companies serving global cloud demand from India.
- Compute-layer cost
- GPUs, servers, networking and liquid cooling commonly account for 60–80% of an AI data centre’s total cost.
- GST rate
- Leasing computer hardware without transferring ownership generally attracts 18% GST in India.
- Indian trust market
- About 32 listed REITs and InvITs hold close to ₹10 lakh crore in assets, with a combined market value of roughly ₹5 lakh crore.
- Proposed financial product
- IFSCA has proposed allowing GPU and related data centre equipment leasing as a financial product in the IFSC.
Sources
Related news
NaBFID Lines Up ₹3,000 Crore Loans to Expand India’s Data Hubs
India Data Centre Capacity Seen Reaching 6 GW by 2029
Nvidia Price Hike Puts India’s AI Data-Centre Build-Out Under Pressure
Avisirah Technologies Raises ₹75 Crore for Navi Mumbai Data Centre
India’s Data Centre Sector Draws Massive Investment Projections
TCS HyperVault Plans ₹70,000-Crore AI Campus in Hyderabad
SpaceX AI Unveils Faster, Cheaper Grok 4.7 Coding Model
Why Frontier AI Development Needs Stronger Government Safeguards
AI Spending Boom Raises Risks for Markets and Economy
AI Agents Expose Cybersecurity Risks as Control Concerns Grow
