Larsen & Toubro has entered the AI factory market with a large-scale infrastructure project in Chennai that will deploy 10,000 NVIDIA B300 GPUs for US-based AI cloud company Together AI.
L&T describes the project as India's largest single-cluster NVIDIA B300 AI Factory. The deployment will be undertaken through Vyoma.AI and its AI infrastructure subsidiary, LTN Compute, and will support Together AI's cloud platform for large-scale AI training, fine-tuning and inference workloads.
The project marks an expansion of L&T's data centre strategy into infrastructure designed specifically for increasingly compute-intensive artificial intelligence workloads.
The AI factory will combine NVIDIA B300 GPUs with high-performance networking, parallel storage, low-latency interconnects and infrastructure operations. According to L&T, the integrated platform will allow customers to deploy and scale AI workloads through a unified infrastructure stack.
L&T has classified the contract as a "mega" order. Under the company's project classification system, mega orders are valued between ₹10,000 crore and ₹15,000 crore.
The infrastructure will be hosted at Vyoma's Chennai data centre campus. The first phase of the site is planned with 250 MW of capacity and power infrastructure readiness of 150 MVA, leaving room for additional AI infrastructure deployments as demand grows.
Together AI provides an AI Acceleration Cloud used by developers, enterprises and AI companies for developing and operating generative AI applications. The Chennai deployment will provide additional computing capacity for workloads running through its platform.
The project is part of L&T's broader Gigawatt AI Infrastructure Mission, through which the engineering conglomerate is building infrastructure for high-density AI computing in India.
Earlier this year, L&T announced plans with NVIDIA to develop sovereign, gigawatt-scale AI factory infrastructure in the country. The initiative combines L&T's engineering and infrastructure capabilities with NVIDIA's accelerated computing stack, including GPUs, CPUs, networking, storage, enterprise AI software and reference architectures.
The companies had outlined plans to support AI workloads across sectors including manufacturing, infrastructure, energy, financial services, healthcare and public services. L&T also said the infrastructure would enable critical data, models and workloads to be built and deployed within India.
Beyond the Together AI deployment, LTN Compute is building an infrastructure portfolio spanning hyperscale AI data centres, sovereign cloud platforms, GPU-as-a-Service, AI factory services and managed AI platforms.
The strategy places L&T in an infrastructure segment that is expanding alongside the adoption of generative and agentic AI. Training and operating increasingly complex AI models requires high-density computing capacity as well as specialised networking, storage, cooling and power infrastructure.
For enterprises and cloud providers, access to such infrastructure is becoming an important part of moving AI projects from experimentation to production.
The 10,000-GPU Chennai deployment gives L&T an anchor project for that strategy while extending its role from conventional data centres into specialised AI computing infrastructure serving both Indian and international customers.