Larsen & Toubro (L&T) announced a strategic venture with NVIDIA on February 18, 2026, to construct gigawatt-scale AI factory infrastructure across India. Operating under the federal IndiaAI Mission, the initiative deploys 30 MW of NVIDIA GPU capacity at L&T's 300-acre Chennai campus and an additional 40 MW at a new Mumbai facility. This deployment targets the transition of Indian enterprises from AI experimentation to production-grade workloads, offering compute at approximately ₹65 per hour—nearly 66% lower than the global average.
The partnership integrates NVIDIA's full-stack accelerated computing—including Blackwell Ultra GPUs, networking, and the AI Enterprise software suite—with L&T's industrial engineering. Beyond public-facing infrastructure, L&T will deploy autonomous AI agents across its group companies, utilizing NVIDIA Omniverse to optimize manufacturing and energy operations. This "sovereign-by-design" model ensures data residency while maintaining global interoperability, positioning India as a strategic hub for global hyperscalers and domestic enterprises.
By controlling the physical layer (L&T) and the compute layer (NVIDIA), this venture creates a vertically integrated alternative to US-based hyperscalers. The strategy directly addresses the "export data to import intelligence" deficit, ensuring that the economic value of Indian data remains within domestic borders.
Nvidia CEO Jensen Huang discusses the strategic value of Indian AI infrastructure
This video provides direct insight into Jensen Huang's perspective on India's role in the global AI ecosystem, reinforcing the "sovereign AI" thesis mentioned in the briefing.
