What Happened
- The India-US interim trade framework includes a provision to "significantly increase trade in technology products, including GPUs and other goods used in data centres," signalling deeper cooperation in advanced technologies.
- The joint statement positions India as the leading democratic alternative to China for high-end artificial intelligence computing, providing a "Strategic GPU Pact" that assures India will not face sudden AI hardware export restrictions similar to those imposed on China.
- Union IT Minister Ashwini Vaishnaw stated that the trade pact combined with Union Budget measures could unlock as much as $200 billion in new data centre investment.
- Import duties of 20-28% on enterprise GPU servers have made GPU-as-a-service pricing in India nearly 40% higher than peer hubs like Singapore and the UAE; duty rationalisation under the deal could reduce data centre setup costs by approximately 14%.
- Major corporate investment announcements followed: Google committed $15 billion for a 1 GW data centre with Adani Group, Microsoft committed $17.5 billion for AI data centres, and Amazon plans $35 billion over five years in India.
Static Topic Bridges
IndiaAI Mission and National Compute Infrastructure
India launched the IndiaAI Mission in March 2024 with a Rs 10,372 crore ($1.25 billion) budget over five years, structured around seven pillars: compute capacity, datasets, innovation centres, application development, future skills, startup financing, and safe and trusted AI. The mission aims to build sovereign AI compute capacity and democratise access.
- Total budget: Rs 10,372 crore over five years (approved March 2024)
- National GPU compute capacity: crossed 34,000 GPUs; target deployment of 38,000 GPUs
- 600 AI data labs planned across the country
- 18,000+ affordable AI compute units available at up to 40% reduced cost under the mission
- AIKosh dataset platform: hosts 7,541 datasets and 273 AI models across 20 sectors (as of February 2026)
- 12 indigenous AI startups selected for model development including Sarvam AI, Gnani AI, BharatGen (IIT Bombay)
- Seven pillars: compute, datasets, innovation, applications, skills, startup financing, safe AI
Connection to this news: The India-US GPU access provisions directly complement the IndiaAI Mission's compute infrastructure goals. While the mission builds sovereign capacity, the trade deal ensures continued access to cutting-edge US-made GPUs (primarily NVIDIA H100/H200 and successor chips), which India cannot manufacture domestically but needs for frontier AI development.
US Export Controls on AI Chips and the Geopolitics of Compute
The United States has imposed escalating export restrictions on advanced AI semiconductors to China since October 2022, targeting high-end GPUs and AI accelerator chips manufactured by NVIDIA, AMD, and Intel. These controls aim to prevent China from acquiring compute capabilities for military AI applications. The restrictions created a tiered global access framework where allied democracies receive preferential treatment.
- October 2022: US Commerce Department Bureau of Industry and Security (BIS) first restricted AI chip exports to China
- October 2023: Updated rules expanded restrictions to include NVIDIA H800 and A800 chips designed to comply with earlier limits
- January 2025: BIS proposed a global tiered framework categorising countries by AI chip access levels
- Tier 1 (unrestricted): US allies including Japan, South Korea, EU members, Australia, UK
- Tier 2 (capped access): most other countries including India
- Tier 3 (restricted): China, Russia, Iran, North Korea
- Key restricted chips: NVIDIA A100, H100, H200, B100 series and equivalents above specified compute thresholds
Connection to this news: The trade deal's "Strategic GPU Pact" provision signals that India may receive treatment closer to Tier 1 allies rather than the capped Tier 2 access. This assurance is critical for India's AI ambitions because without guaranteed access to frontier GPUs, both government AI missions and private sector data centre investments would face supply uncertainty.
India's Data Centre Industry and Infrastructure Growth
India's data centre capacity has grown rapidly, driven by digital transformation, cloud adoption, and the government's data localisation policies. India currently has approximately 1.1-1.3 GW of installed IT load capacity across major hubs in Mumbai, Chennai, Hyderabad, Pune, and Delhi-NCR. The industry is projected to grow significantly, with estimates of 3-5 GW capacity by 2028.
- Current installed IT load capacity: approximately 1.1-1.3 GW
- Primary data centre hubs: Mumbai (largest, ~50% of capacity), Chennai, Hyderabad, Pune, Delhi-NCR
- Major operators: NTT, Equinix, STT Telemedia, Adani Group, Reliance (Jio)
- Union Budget 2025-26: Infrastructure status for data centres, enabling access to long-term finance
- Data localisation requirements: RBI mandated payment data storage in India (2018); Personal Data Protection Act (2023) provisions
- Power consumption: a major constraint; typical data centre uses 5-10 MW, hyperscale facilities use 50-200 MW
- Estimated investment needed by 2030: $35-50 billion for data centre infrastructure
Connection to this news: The announced corporate investments (Google $15 billion, Microsoft $17.5 billion, Amazon $35 billion) represent a potential tripling of India's data centre capacity. The trade deal's duty rationalisation on GPU servers and data centre equipment directly lowers the capital expenditure required, making India more competitive as a regional data centre hub compared to Singapore and the UAE.
Semiconductor Supply Chain and India's Chip Ambitions
India has launched the India Semiconductor Mission (ISM) with a Rs 76,000 crore ($10 billion) outlay to build domestic semiconductor manufacturing capacity. While India does not currently manufacture advanced logic chips, it has a significant presence in chip design, with over 20% of the world's semiconductor design engineers based in India.
- India Semiconductor Mission (ISM): Rs 76,000 crore budget
- Tata Electronics-PSMC fab in Gujarat (Dholera): 28nm process, construction underway, expected operational by 2026-27
- Micron ATMP (assembly, testing, packaging) facility in Gujarat (Sanand): $2.75 billion investment
- CG Power (Renesas partnership): proposed OSAT facility
- India's semiconductor design workforce: over 20% of global chip design engineers
- India's semiconductor import bill: approximately $15-20 billion annually
- No domestic production of advanced AI GPUs; entirely dependent on imports (NVIDIA, AMD)
Connection to this news: Until India develops domestic semiconductor fabrication for advanced chips (which remains years away even for 28nm, let alone the 5nm/4nm nodes used in AI GPUs), the trade deal's guaranteed GPU access provision is essential. It bridges the gap between India's current import dependence and its long-term semiconductor manufacturing ambitions under ISM.
Key Facts & Data
- GPU server import duties (pre-deal): 20-28%, making India 40% costlier than Singapore/UAE for GPU-as-a-service
- Estimated cost reduction from duty rationalisation: approximately 14% for GPU-ready data centres
- Potential data centre investment unlocked: $200 billion (per IT Minister Vaishnaw)
- Google investment: $15 billion for 1 GW data centre (with Adani Group)
- Microsoft investment: $17.5 billion for AI data centres
- Amazon investment: $35 billion over five years in India
- IndiaAI Mission budget: Rs 10,372 crore ($1.25 billion) over five years
- National GPU compute capacity: 34,000+ GPUs deployed; 38,000 target
- India Semiconductor Mission: Rs 76,000 crore ($10 billion) outlay
- US reciprocal tariff on India: reduced from 25% to 18%
- India's $500 billion five-year purchase commitment includes technology products