India's AI infrastructure race is getting a major new entrant.
Larsen & Toubro (L&T) has secured a mega order tied to the construction of what the company describes as India's largest single-cluster AI infrastructure, a massive NVIDIA B300-powered AI Factory that will be hosted at its Chennai data centre campus.
According to L&T's official announcement, the facility will have capacity for 10,000 NVIDIA B300 GPUs and will power the AI-native cloud platform of US-based Together AI.
The project isn't simply another conventional data centre. It is being designed specifically for demanding artificial intelligence workloads including large-scale model training, fine-tuning and inference.
And its scale makes the announcement particularly significant for India's ambitions to build more AI computing capacity domestically.
A Mega AI Infrastructure Deal
The project is being undertaken through LTN Compute, the AI infrastructure subsidiary of L&T's Vyoma.AI business.
L&T classifies contracts valued between ₹10,000 crore and ₹15,000 crore as "Mega" orders, although the company did not disclose an exact contract value in its announcement.
Reuters reported that the order could be worth up to ₹150 billion, or ₹15,000 crore, and noted that L&T shares rose as much as 1.2% during Thursday trading following the announcement.
More important than the headline value, however, is what L&T is actually building.
The company says the integrated AI Factory will combine hyperscale data-centre infrastructure with accelerated computing, high-performance networking, ultra-low-latency interconnects, high-throughput parallel storage and AI infrastructure operations.
In simpler terms, the facility is being designed as a complete computing environment where extremely demanding AI models can be trained and operated at scale.
10,000 NVIDIA B300 GPUs Are at the Centre of the Project
The headline hardware is NVIDIA's B300 platform.
L&T says its Chennai AI Factory will have capacity for 10,000 NVIDIA B300 GPUs, creating an enormous pool of accelerated computing power for Together AI.
According to the company's official project announcement, those GPUs will support three particularly important categories of AI workloads: training, fine-tuning and inference.
Training is the computationally intensive process used to create or further develop AI models from large datasets.
Fine-tuning adapts an existing model for particular tasks, industries or datasets.
Inference is what happens when a trained AI model actually processes a request and generates an output.
That final category is becoming particularly important as generative AI services gain more users.
Reuters recently reported that growing use of AI inference has become a major driver of computing demand, pushing cloud providers and chipmakers to expand AI infrastructure.
The Chennai project is therefore arriving as demand shifts from simply training ever-larger models toward running those models at enormous scale.
Why Together AI Needs So Much Computing Power
Together AI is an AI cloud company that provides infrastructure, models and developer services for companies building generative AI applications.
The company needs significant GPU capacity because modern AI applications can require huge amounts of accelerated computing for both development and everyday operation.
The Chennai project also isn't Together AI's only major infrastructure expansion.
Just two days before L&T's announcement, Reuters reported that Together AI and IBM had signed a $240 million multi-year agreement for a large-scale AI inference cluster on IBM Cloud using NVIDIA systems.
That planned initial cluster uses 2,000 NVIDIA B300 chips.
By comparison, L&T's Chennai AI Factory is designed for capacity of 10,000 B300 GPUs.
The comparison gives some perspective on just how large the Indian deployment could become.
Chennai Is Becoming the Home of the AI Factory
The AI Factory will be hosted at Vyoma.AI's Chennai data centre campus.
The location itself is being designed for much more than this single deployment.
L&T says the Chennai campus is a gigawatt-scale AI infrastructure site, with its first phase designed for 250 MW and power infrastructure readiness of 150 MVA, providing room for future AI Factory expansion.
That power requirement highlights an important reality of modern AI.
GPUs receive most of the attention, but building large-scale AI infrastructure requires much more than purchasing processors.
Thousands of high-performance accelerators need enormous amounts of electricity, cooling, networking bandwidth and storage. They also need infrastructure capable of keeping those systems running reliably around the clock.
That is where a company with L&T's traditional engineering and infrastructure experience could have an advantage.
L&T Is Moving Beyond Traditional Data Centres
For L&T, the deal represents something bigger than another construction contract.
It marks the company's entry into the AI Factory business.
Through LTN Compute, L&T says it is developing AI-ready infrastructure across India that includes hyperscale AI data centres, sovereign cloud platforms, AI Factory services, GPU-as-a-Service and managed AI platforms.
That represents a significant expansion of what a traditional infrastructure company can provide.
Instead of simply constructing the physical building where servers are installed, L&T is moving toward providing an integrated computing platform encompassing power, networking, storage and GPU infrastructure.
L&T Chairman and Managing Director S N Subrahmanyan said in the company's announcement that AI is becoming foundational across industries and described the Together AI deployment as a milestone in L&T's Gigawatt AI Infrastructure Mission.
Why This Matters for India's AI Ambitions
India has plenty of AI developers, startups and technology companies.
But developing advanced AI services requires something much more physical: computing infrastructure.
Access to powerful GPUs has become one of the critical resources in the global AI industry. Companies building large models need clusters containing thousands of accelerators, while businesses deploying those models need increasingly large amounts of inference capacity.
Having more of that infrastructure inside India could provide several strategic advantages.
Domestic GPU capacity can support Indian startups and enterprises without making them entirely dependent on computing infrastructure located overseas.
It could also help businesses with workloads where data location, latency or regulatory requirements matter.
And if India develops enough high-performance capacity, those data centres don't have to serve only Indian customers. Projects such as Together AI's deployment show that infrastructure located in India can potentially support global AI workloads as well.
That changes the conversation from India simply using AI products to India increasingly becoming one of the places where the infrastructure powering those products is located.
The Bigger Story Is India's AI Infrastructure Build-Out
The most eye-catching number in L&T's announcement is undoubtedly 10,000 NVIDIA B300 GPUs.
But the number alone isn't the most important part of the story.
The bigger development is that Indian infrastructure companies are beginning to build facilities specifically designed around artificial intelligence rather than treating AI workloads as another type of conventional cloud computing.
L&T's combination of a gigawatt-scale Chennai campus, high-density NVIDIA computing infrastructure and Together AI's cloud platform illustrates how quickly that transition is happening.
For developers, more infrastructure could eventually mean greater access to high-performance computing.
For enterprises, it could create additional options for running AI workloads closer to Indian operations and data.
And for India's technology industry, it creates an opportunity to participate in another layer of the AI economy — not only applications and software, but the enormous physical computing infrastructure underneath them.
The Chennai facility won't by itself determine India's position in the global AI race.
But a 10,000-GPU AI Factory is a strong indication that the country's next phase of AI growth will be measured not just in models and apps, but also in megawatts, data centres and accelerated computing capacity.