3 min readFeb 17, 2026 07:11 AM IST
First published on: Feb 17, 2026 at 07:11 AM IST
India’s trade deal with the United States marks a significant moment in its AI ambitions, expanding access to graphics processing units, deepening technology cooperation and encouraging massive foreign investment in data centres. Alongside this, the Union Budget’s generous tax holidays for foreign hyperscalers signal that India is betting heavily on building its AI economy.
The short-term gains are undeniable. India lacks domestic GPU manufacturing capacity and remains heavily dependent on imports dominated by US firms. The trade framework helps ease these constraints by expanding access to high-end chips and helping India escape export-control restrictions that had capped imports of advanced GPUs. This is critical because AI innovation today is shaped largely by computing power. Although the current GPU capacity under the IndiaAI Mission stands at about 40,000, far below that of leading global companies, the government is betting that this number will rise to 100,000 by the end of the year.
India has also rolled out sweeping fiscal incentives to attract global cloud providers. Analysts estimate that the incentives in the Budget could catalyse tens of billions of dollars in data-centre expansion while lowering compute costs for start-ups and enterprises.
However, these policies deepen structural dependence on foreign technology-providers. India is offering long-term fiscal concessions and market access without clear guarantees of technology transfer or domestic intellectual property creation. It has indicated that it intends to purchase around $500 billion worth of US goods over five years, including technology products and chips. At the same time, allocation for the IndiaAI Mission has been reduced from Rs 2,000 crore to Rs 1,000 crore. Experts have flagged the absence of sustained frontier research funding and long-term innovation incentives as a major vulnerability.
Rapid data-centre expansion raises environmental and democratic concerns. A hyperscale facility near Greater Noida is expected to consume about 160 megawatts of electricity, while global estimates suggest that even a one-megawatt data centre can use up to 25.5 million litres of water annually. In water-stressed regions, this creates competition between AI infrastructure and needs of communities, who often learn about projects only after construction begins.
To be fair, India’s strategy reflects pragmatic sequencing. Building domestic compute ecosystems requires enormous capital and expertise. Leveraging foreign investment and global supply chains may be necessary to build early capacity. This year’s Budget treats AI as foundational infrastructure. Yet, infrastructure alone does not guarantee technological sovereignty or equitable development. If foreign hyperscalers dominate India’s data ecosystem while domestic research funding stagnates, India risks becoming only a hosting ground. If environmental costs and community displacement remain externalised, AI infrastructure could replicate earlier development models where economic gains were national but ecological and social burdens remained local.
The India-US deal is a strategic gamble. It delivers compute access and investment momentum, but without stronger domestic innovation investments, environmental safeguards and democratic participation, India may find that its AI ambitions are built on foundations that are economically productive but socially and ecologically fragile.
The writer, a former fellow of Harvard Kennedy School, teaches at Jindal School of Government and Public Policy
