Larsen & Toubro (L&T) says a planned AI factory at its Chennai campus marks its entry into the business, with capacity expected to exceed 10,000 NVIDIA B300 graphics processing units (GPUs).
At L&T’s Oct. 27 Q2 and first-half FY2027 earnings call, Chief Financial Officer Parameswaran Ramakrishnan set January 2027 as the commissioning target. He described capacity above 10,000 GPUs, while the company’s Aug. 13 announcement specified an order for 10,000. L&T did not explain the difference between the two figures.
Ramakrishnan said LTN Compute has a long-term deployment arrangement with Together AI, which is expected to use a substantial share of the factory’s capacity. L&T said the facility would support training and inference work on Together AI’s cloud platform. In its August announcement, L&T Chairman and Managing Director S. N. Subrahmanyan described the project as part of the company’s “Gigawatt AI Infrastructure Mission.”
Discussing the GPU plan and an existing 30-megawatt data center near Chennai, Ramakrishnan cited about $750 million in investment but gave no detailed cost breakdown. He also outlined a 50-megawatt AI data center beside the existing center in Sriperumbudur, plus another 30 megawatts.
L&T’s August announcement put the broader Vyoma.AI Chennai campus’s Phase 1 design capacity at 250 megawatts and said 150 MVA of power infrastructure was ready. It did not explain how those campus specifications relate to the GPU and AI data-center figures discussed on the call.
Ramakrishnan said L&T aims to exceed 100 megawatts of AI data-center capacity over the next 1.5 to two years.
