• L&T's Vyoma.AI business will deploy a 10,000-GPU Nvidia B300 AI Factory for Together AI at its Chennai data-centre campus
  • Phase one of the campus is designed for 250MW, while L&T cites 150MVA of power-infrastructure readiness without giving a commissioning timetable

The fact

Larsen & Toubro's Vyoma.AI business, through subsidiary LTN Compute, has secured an order from Together AI to deploy an Nvidia B300 AI Factory at its Chennai data-centre campus. L&T says the cluster will have capacity for 10,000 B300 GPUs and support Together AI's cloud platform for training, fine-tuning and inference.

The deployment combines data-centre infrastructure, accelerated computing, high-performance networking, low-latency interconnects, parallel storage and infrastructure operations. L&T says the Chennai campus is being developed at gigawatt scale, with its first phase designed for 250MW and 150MVA of power-infrastructure readiness.

L&T classifies the order as “Mega”, which under its published classification covers ₹10,000 crore to ₹15,000 crore. It has not disclosed an exact contract value, commissioning date, final cluster power requirement or GPU delivery timetable.

The assessment

The order gives L&T responsibility for more than the data-centre shell. Vyoma is expected to provide the facility together with the accelerators, networking, storage and infrastructure operations needed for Together AI to run its cloud platform.

That means the cluster cannot enter service until several systems are ready at the same time. GPU installation must be matched by usable power, cooling, network capacity and storage. L&T's 250MW phase-one design indicates the scale planned for the Chennai campus, but it does not show how much capacity is already energised or how much will support this particular deployment. The same uncertainty applies to the stated 150MVA of power-infrastructure readiness.

For BTW readers, the next test is whether L&T can turn the 10,000-GPU order into a commissioned cluster that Together AI can use. GPU deliveries, energised capacity and completion of the supporting network, storage and cooling systems will show how far the project has moved from contract to operation.

What to watch

Watch for the first GPU deliveries, energised capacity allocated to the cluster, completion of its network and storage infrastructure, and a date for Together AI to begin using it. Disclosure of the cluster's final IT load and cooling configuration would also show how the 10,000-GPU deployment fits within the larger 250MW first phase.