India Eyes Sovereign GPUs with NVIDIA

India’s push toward technological sovereignty took a significant step forward as Union Minister for Electronics and Information Technology Ashwini Vaishnaw engaged in high‑level discussions with NVIDIA executives on local GPU manufacturing and advanced AI hardware production. Building on the success of Apple’s local assembly, the talks signal a broader ambition to reduce reliance on imported semiconductor technology and strengthen India’s position in the global AI ecosystem.

The centerpiece of the dialogue was NVIDIA’s newly unveiled DGX Spark, a compact AI supercomputer showcased at CES 2026. Designed for edge computing, Spark delivers up to 1 petaFLOP of performance and supports large language models with as many as 200 billion parameters. Crucially, it operates fully offline, enabling secure AI inference in sensitive or remote environments such as railways, shipping, healthcare, and education. Vaishnaw emphasized that such devices could play a transformative role in India’s digital infrastructure, aligning with the Atmanirbhar Bharat initiative.

India’s long‑term goal is to develop sovereign GPU designs within three to four years, potentially leveraging open‑source architectures or licensed technologies. This ambition is reinforced by the India AI Mission, which has already deployed 38,000 subsidized GPUs at affordable rates and selected 12 startups to build indigenous AI models.

Global demand for AI chips is surging, and supply chains remain vulnerable to geopolitical risks. With NVIDIA controlling over 80% of the GPU market, partnerships of this nature could accelerate India’s entry into high‑end semiconductor manufacturing. While no immediate agreements were announced, the discussions underscore India’s determination to build a domestic AI hardware ecosystem capable of supporting both national security and economic growth.

By marrying policy vision with industry collaboration, India is positioning itself to become not just a consumer of AI technology, but a producer of the hardware that powers it.