A high-level Indian delegation, including foreign secretary Vikram Misri and deputy NSA Pavan Kapoor, met their US counterparts in Washington this week. Among other issues, they discussed implementing TRUST (Transforming the Relationship Utilising Strategic Technology) initiative, and US-India COMPACT (Catalysing Opportunities for Military Partnership, Accelerated Commerce & Technology) partnership. At the centre of Misri's discussions with US deputy secretary of state Christopher Landau was how tech-trade-talent will shape the India-US partnership.
In May, the Trump administration had repealed the AI Diffusion Rule introduced towards the end of Joe Biden's tenure. The rule classified countries across three tiers, laying down specific criteria for exporting - or not - AI infrastructure to them. India, placed in tier 2, with about 120 other countries, faced a cap of 50,000 advanced GPUs to be imported from the US through 2027. Also, no US company could deploy over 7% of its overall compute power in India.
That rule is slated to be replaced, with recent reports suggesting that new rules on exporting US AI infra could be driven by country-specific negotiations, and may feature in their ongoing bilateral trade negotiations. For India, there are no clear indications yet on what the new rules could mean for it. As of now, repeal of AI Diffusion Rule is widely seen as a 'big breather'.
However, importing an uncapped number of GPUs, or other advanced AI chips, from the US may not necessarily translate to sufficient compute capacity for India. AI Diffusion Rule's repeal should be an opportunity to understand other roadblocks to building compute capacity for India.
As computing power is necessary to fuel any AI ambitions, GoI recently provisioned about 14,000 GPUs on subsidy for its AI ecosystem. It's expected to procure 15,000 more from 7 shortlisted bidders in the coming months.
Regardless of volume, any GPUs or advanced AI chips imported need to be housed in data centres. This assumes significance as a May 2025 Deloitte report, 'Attracting AI Data Centre Infrastructure Investment in India', highlights that India may need 45-50 mn sq ft real estate, coupled with 40-45 terawatt hours (TWh) incremental power for data centres by 2030 for its AI demand. So, challenges to building data centres must remain in India's policymaking focus.
The rule repeal remains positive news, though limited in its impact and contingent on other factors relating to India's overall compute capabilities. But it serves as an opportunity to clear the mist on capabilities of India's data centres that will ultimately house high-end AI chips from the US. Both governments must now double down on identifying roadblocks to bringing AI infra to India, particularly under the timely TRUST initiative.
Further, the Quad leaders' summit, to be hosted by India later this year, can be opportune platform for presenting Washington's and New Delhi's findings on TRUST. This could lead to reformative policymaking relating to power, finance and regulatory roadblocks.
Compute capacity is only one of the building blocks for India's AI ambitions. Launched last year, the IndiaAI Mission, with a ₹10,300 cr-plus budget outlay, has made significant strides on its seven pillars. In addition to its massive and ongoing compute procurement, its AIKosh platform, unveiled in March, now boasts 350+ datasets and will only grow.
Last month, GoI selected the Bengaluru-based startup Sarvam AI to build India's first sovereign LLM model. It also awarded several projects for safety-enhancing technologies, and has called for partnerships for its AI Safety Institute. But more needs to be done, especially to address AI-induced structural unemployment, and the need to increase public-private spending on the R&D ecosystem.
The writer is associate director,US-India Strategic Partnership Forum
In May, the Trump administration had repealed the AI Diffusion Rule introduced towards the end of Joe Biden's tenure. The rule classified countries across three tiers, laying down specific criteria for exporting - or not - AI infrastructure to them. India, placed in tier 2, with about 120 other countries, faced a cap of 50,000 advanced GPUs to be imported from the US through 2027. Also, no US company could deploy over 7% of its overall compute power in India.
That rule is slated to be replaced, with recent reports suggesting that new rules on exporting US AI infra could be driven by country-specific negotiations, and may feature in their ongoing bilateral trade negotiations. For India, there are no clear indications yet on what the new rules could mean for it. As of now, repeal of AI Diffusion Rule is widely seen as a 'big breather'.
However, importing an uncapped number of GPUs, or other advanced AI chips, from the US may not necessarily translate to sufficient compute capacity for India. AI Diffusion Rule's repeal should be an opportunity to understand other roadblocks to building compute capacity for India.
As computing power is necessary to fuel any AI ambitions, GoI recently provisioned about 14,000 GPUs on subsidy for its AI ecosystem. It's expected to procure 15,000 more from 7 shortlisted bidders in the coming months.
Regardless of volume, any GPUs or advanced AI chips imported need to be housed in data centres. This assumes significance as a May 2025 Deloitte report, 'Attracting AI Data Centre Infrastructure Investment in India', highlights that India may need 45-50 mn sq ft real estate, coupled with 40-45 terawatt hours (TWh) incremental power for data centres by 2030 for its AI demand. So, challenges to building data centres must remain in India's policymaking focus.
The rule repeal remains positive news, though limited in its impact and contingent on other factors relating to India's overall compute capabilities. But it serves as an opportunity to clear the mist on capabilities of India's data centres that will ultimately house high-end AI chips from the US. Both governments must now double down on identifying roadblocks to bringing AI infra to India, particularly under the timely TRUST initiative.
Further, the Quad leaders' summit, to be hosted by India later this year, can be opportune platform for presenting Washington's and New Delhi's findings on TRUST. This could lead to reformative policymaking relating to power, finance and regulatory roadblocks.
Compute capacity is only one of the building blocks for India's AI ambitions. Launched last year, the IndiaAI Mission, with a ₹10,300 cr-plus budget outlay, has made significant strides on its seven pillars. In addition to its massive and ongoing compute procurement, its AIKosh platform, unveiled in March, now boasts 350+ datasets and will only grow.
Last month, GoI selected the Bengaluru-based startup Sarvam AI to build India's first sovereign LLM model. It also awarded several projects for safety-enhancing technologies, and has called for partnerships for its AI Safety Institute. But more needs to be done, especially to address AI-induced structural unemployment, and the need to increase public-private spending on the R&D ecosystem.
The writer is associate director,US-India Strategic Partnership Forum
(Disclaimer: The opinions expressed in this column are that of the writer. The facts and opinions expressed here do not reflect the views of www.economictimes.com.)
Shatakratu Sahu
Shatakratu Sahu is associate director, US-India Strategic Partnership Forum.