India Commits ₹10,371 Crore to AI Mission, Deploys 38,000 GPUs in Data Sovereignty Push

Brian Roemmele, a prominent AI expert, recently issued a stark warning regarding a foreign government's aggressive AI strategy, stating, "Government supplied FREE training data, because that government actually has a Manhattan Project to preserve non-internet data is propelling this country’s AI rapidly." This observation aligns with India's recently unveiled "BharatGen AI" initiative, a government-funded project dedicated to building multimodal large language models using extensive domestic datasets, backed by a significant ₹10,371 crore investment over five years. The IndiaAI Mission has already deployed 38,000 GPUs to support this ambitious endeavor.

Roemmele has consistently voiced concerns about the quality degradation of internet-sourced data for AI training and the strategic importance of offline data. He has highlighted the necessity for nations to cultivate their own robust, diverse datasets to achieve AI leadership, noting that the United States risks falling behind without similar initiatives. His work, including advancements in processing offline data with tools like DeepSeek-OCR, underscores the technical feasibility and strategic value of such data.

The BharatGen AI initiative, launched on June 2, 2025, is spearheaded by IIT Bombay under the National Mission on Interdisciplinary Cyber-Physical Systems. It aims to create generative AI systems capable of producing high-quality text and multimodal content in 22 Indian languages. This project explicitly focuses on "collecting and curating India-centric data," ensuring that the nation's diverse linguistic and cultural contexts are accurately represented, thereby fostering data sovereignty.

This strategic move is a core component of the broader IndiaAI Mission, which seeks to establish India as a global leader in artificial intelligence by reducing reliance on foreign technologies. The mission emphasizes developing foundational AI models tailored for India, democratizing access to AI, and fostering a vibrant domestic AI research community through training programs and academic collaborations. The significant investment and infrastructure deployment signal a clear national priority.

Globally, many countries are developing national AI strategies, but India's explicit focus on government-funded, homegrown, multilingual, and multimodal models built on domestic datasets sets it apart. This approach directly addresses the kind of "Manhattan Project" for non-internet data that Roemmele described as a key accelerator for national AI capabilities. Such initiatives are seen as crucial for rapid advancement in the competitive global AI landscape.