Inside L&T Vyoma: India's New AI Compute Giant

Larsen & Tourbo isn't the name most people associate with artificial intelligence. For decades, it's been the company that built India's ports, power plants, metros and refineries. But over the past year, the engineering conglomerate has been assembling a completely different kind of infrastructure business — one measured not in tonnes of steel or kilometres of pipeline, but in megawatts of GPU compute.

That business now has a name: Vyoma.AI. And its ambitions are enormous.

From Construction Major To AI Infrastructure Player

L&T's pivot into AI infrastructure didn't happen overnight. It began earlier this year with a partnership with NVIDIA to build a gigawatt-scale AI data centre in India, announced at the India AI Summit under the IndiaAI Mission. The idea was simple in concept but massive in execution: combine L&T's decades of engineering and construction expertise with NVIDIA's compute, networking and software stack to build AI capacity that could serve both domestic enterprises and global hyperscalers.

That partnership evolved into Vyoma.AI Ltd, incorporated as a wholly owned L&T subsidiary with an initial authorised capital of just ₹5 lakh — a modest starting point for what has since become one of the most closely watched infrastructure bets in Indian tech. L&T has since folded its existing data centre and cloud services operations into Vyoma.AI in a ₹1,400 crore internal transfer, consolidating everything under one roof.

What Exactly Is Vyoma Building?

Vyoma describes itself as a next-generation sovereign AI cloud and digital infrastructure platform, built around three pillars: hyperscale cloud engineering, GPU computing, and AI workload management. Its Sovereign Cloud Platform (SCP) is designed around a "jurisdiction-first" architecture — meaning enterprise data, models and AI workloads stay within Indian regulatory boundaries, a pitch aimed squarely at government bodies, banks, and regulated industries wary of routing sensitive data through foreign clouds.

Underneath Vyoma sits LTN Compute, the arm actually responsible for building and operating the physical AI infrastructure — data centres, GPU clusters, high-throughput storage, and the ultra-low-latency networking that ties it all together.

The Chennai Mega-Deal: India's Largest Single AI Cluster

The clearest signal of Vyoma's scale ambitions came with a mega order to build an NVIDIA B300-based "AI Factory" at its Chennai data centre campus — a deal that will power the AI-native cloud platform of US-based Together AI. L&T hasn't disclosed the exact contract value, but under its own project classification, "mega orders" fall in the ₹10,000–15,000 crore range. The specs are striking:

• 10,000 NVIDIA B300 GPUs integrated into a single AI Factory cluster — reportedly India's largest single-cluster AI deployment to date

• Phase 1 capacity of 250 MW, with power infrastructure readiness of 150 MVA

• Designed from the ground up as a gigawatt-scale campus, meaning Chennai's current build is just the opening phase of something much larger

The platform is built to handle large-scale inference, fine-tuning and full model training — essentially the entire lifecycle of enterprise AI workloads — through what Vyoma calls a unified, end-to-end AI infrastructure stack combining compute, networking and storage.

Together AI, notably, has been on an infrastructure shopping spree globally, having also signed a $240 million multi-year deal with IBM for a large-scale AI cluster on IBM Cloud. Landing a partner with that kind of appetite is a strong early validation for Vyoma's global ambitions.

Beyond Chennai: Dholera And Navi Mumbai

Chennai isn't Vyoma's only bet. In Gujarat, L&T Vyoma has signed an agreement with the state government to build a green AI data centre in the Dholera Special Investment Region — a ₹25,000 crore investment for a 250 MW facility powered by clean energy, expected to go live by 2028. The Gujarat government has backed the push with over ₹850 crore earmarked for AI and digital services in its budget.

Vyoma is also building a 100 MW campus in Navi Mumbai, purpose-built for build-to-suit, enterprise and "neocloud" workloads — designed to serve AI-native startups, banks, retailers, healthcare players and government clients alike.

E2E Strategic Investment

L&T is also building out its AI infrastructure business through strategic investments and partnerships. 

In 2024, the company announced an investment of approximately ₹1,407 Cr to acquire a minority stake of up to 21% in listed AI and hyperscale cloud platform provider E2E Networks.

The partnership combines L&T’s data centre footprint with E2E’s GPU-powered cloud infrastructure. E2E is also operating the 2,000 B200 GPUs deployed at the Chennai facility.  GPUs are being operated by E2E for client workloads, while the larger 10,000-GPU B300 deployment will be operated by Vyoma as its own AI factory.

Why L&T Is Betting Big On "Neoclouds"

Seema Ambastha, CEO of L&T's data centre business, has been vocal about why she believes independent AI cloud providers — often called neoclouds — have real staying power against the hyperscaler giants. Her reasoning: incumbents like the major global cloud providers already have deep enterprise relationships that are hard to dislodge for standard public cloud workloads, but AI compute is a different game entirely, driven by raw GPU availability rather than existing customer lock-in.

Her projection is bold — neoclouds could capture 5–10% of the overall cloud market and as much as 30% of AI and GPU-specific workloads over the next three to seven years, with an even more aggressive 25–35% share of India's AI compute market by 2030.

The Bigger Picture: India's AI Infrastructure Gold Rush

Vyoma isn't operating in isolation. It's entering a market where nearly every major Indian conglomerate is racing to stake a claim. Reliance has unveiled a ₹10 trillion (roughly $110 billion) plan to build AI computing infrastructure over the next seven years, including gigawatt-scale data centres and a nationwide edge computing network tied to Jio. Adani Group has outlined roughly $100 billion in AI data centre investment. Tata Group has partnered with OpenAI on a project starting at 100 MW and scaling toward a full gigawatt. Collectively, the Indian Government expects more than $200 billion in AI infrastructure spending over the next two years.

Against that backdrop, India's data centre market itself is projected to reach $11.6 billion by 2032, growing at nearly 11% annually, according to the Economic Survey 2024–25.

What It Means

For L&T, Vyoma represents something bigger than a new business line — it's a hedge against the cyclical nature of construction and core engineering, a bet on high-margin, recurring technology revenue, and an attempt to reposition a century-old industrial major as a serious infrastructure player in the AI era. Whether Vyoma can convert its early mega-deals into a durable, profitable AI cloud business — while competing against Reliance, Adani, Tata-OpenAI and global hyperscalers for the same GPUs, power and customers — will be the real test over the next few years.

What's clear for now is that L&T isn't treating AI infrastructure as a side project. Chennai, Dholera and Navi Mumbai together sketch out a company trying to become one of India's foundational AI compute providers — not just another contractor building someone else's data centre.