Newsletter
Larsen & Toubro is making a serious push beyond traditional infrastructure. The engineering giant has secured a “mega” order to build India’s largest single-cluster AI infrastructure deployment for US-based AI cloud company Together AI.
The order falls in L&T’s ₹10,000-15,000 crore classification and will see its AI infrastructure subsidiary, LTN Compute, deploy an AI factory powered by 10,000 NVIDIA B300 GPUs at L&T Vyoma’s Chennai data centre campus.
What exactly is L&T building: think of an AI factory as a data centre specifically designed to do the heavy computational work behind artificial intelligence.
Regular data centres largely store information, host applications and run websites. AI infrastructure needs significantly more computing power, high-speed networking and specialised chips because thousands of GPUs may need to work together on the same AI workload.
L&T’s facility will combine 10,000 NVIDIA B300 GPUs with high-performance networking, ultra-low-latency connections and high-throughput storage. It will be used by Together AI for three major jobs: training AI models, fine-tuning existing models and inference, which is essentially running trained AI models to generate answers or predictions.
NVIDIA’s B300 platform is designed for demanding AI workloads, making the project substantially different from simply constructing another commercial data centre.
Why Together AI needs L&T: Together AI operates an AI-native cloud platform.
In simple terms, companies building AI applications can rent computing infrastructure from Together AI rather than buying and operating thousands of expensive GPUs themselves.
L&T brings something different to the table: the ability to build and operate the physical infrastructure required to run those chips at enormous scale.
And that means much more than putting GPUs into racks. Large AI clusters need enormous amounts of electricity, cooling, storage and extremely fast connections between processors.
The Chennai campus has been planned as a gigawatt-scale AI infrastructure site. Its first phase is designed for 250 MW, with 150 MVA of power infrastructure readiness, leaving room for considerably larger deployments later.
This is part of a much bigger Vyoma.AI bet
Earlier this year, L&T announced a partnership with NVIDIA to develop gigawatt-scale AI factory infrastructure in India. The plan included scaling GPU clusters at its Chennai facility and expanding AI-ready capacity in Mumbai.
L&T Vyoma has also launched a Sovereign Cloud Platform, targeting enterprises and government organisations that want sensitive data, AI models and workloads to remain under greater domestic control.
Meanwhile, construction is underway on a 40 MW green, AI-ready data centre at Mahape in Navi Mumbai, forming the first stage of a campus that can eventually scale to 100 MW.
AI is creating a new category of infrastructure: instead of only building roads, metros and industrial plants, L&T now wants to build the physical backbone on which AI models run.
The Together AI contract is important because it gives that strategy a large commercial customer. Rather than merely announcing future AI capacity, L&T has now landed a mega order that could be worth as much as ₹15,000 crore.
And the 10,000-GPU deployment may only be the beginning. The Chennai site's much larger power capacity gives L&T room to add significantly more AI computing infrastructure if demand follows.
