Build, Not Buy: The Semiconductor Lesson that Indian AI Experts Must Learn

Carnegie India  |  Konark Bhandari

The India AI Mission, despite assembling 38,000 graphics processing units by late 2025, faces severe strategic vulnerabilities due to its heavy reliance on imported hardware and foreign frontier models. This dependency exposes the domestic ecosystem to sudden geopolitical disruptions, such as United States export controls and restricted access to advanced models like Anthropic’s Claude Mythos.

Historically, national technology strategies have underestimated the risks of single-vendor supply chains, assuming global access to computational infrastructure would remain permanently open and unhindered. However, rising technology nationalism and strict export regimes threaten to inflate the costs of digital public infrastructure, rendering application-only strategies economically unsustainable. To mitigate these vulnerabilities, domestic policymakers are shifting focus toward advanced packaging and high-bandwidth memory to secure self-reliance. By building indispensable hardware capabilities rather than merely procuring foreign compute, the nation aims to establish leverage within the contested global technology supply chain.

Comment

The Tata-PSMC semiconductor fabrication facility in Dholera anchors India's transition from a software-first economy to a hardware manufacturer. This domestic manufacturing capability reduces New Delhi's dependence on imported graphics processing units. Localising physical infrastructure shields Indian artificial intelligence initiatives from unilateral export restrictions like the US Framework for AI Diffusion. Consequently, the Dholera facility establishes a foundation for sovereign compute capacity that cannot be easily choked by external geopolitical actors.

To read in details click here