India bets on sovereign AI with a GPU-driven upgrade

The Ministry of Electronics and Information Technology (MeitY) has taken a concrete step toward a self-reliant AI future by approving a landmark contract. E2E Networks, a Bengaluru-based cloud provider backed by Larsen & Toubro, will supply a high-performance GPU infrastructure to Gnani.ai as part of the IndiaAI Mission funding. Think of it as laying down the rails for a domestic high-speed train that’s designed to run on Indian tracks, built by Indian engineers, for Indian passengers—in other words, AI models trained to work with India’s languages, cultures, and data realities.

What’s on the hardware slate?

The contract isn’t just about more graphics cards. It envisions a tightly integrated HPC environment, featuring NVIDIA’s H100 SXM and the newer H200 SXM GPUs. These chips aren’t simply powerful; they’re designed to accelerate the kind of large-scale model training that used to be the exclusive purview of global hyperscalers. And this setup isn’t standalone: the GPUs will be connected via an InfiniBand network on a single fabric. That matters because, in the world of large AI models, latency and bandwidth can make or break a training run.

• A single, cohesive HPC fabric to minimize data movement bottlenecks
• H100 SXM and H200 SXM GPUs designed for large-scale model training
• A focus on ultra-low latency and high-throughput interconnects

Over 360 days, E2E will deliver 12,994,560 GPU hours. That figure isn’t just a marketing number; it maps to the sustained compute needed to push a large foundational model from scratch. It’s the kind of resource that, for many startups and researchers, would otherwise require a multi-cloud, multi-vendor procurement—often with price tags that make ambitious experiments feel like a luxury.

“It’s not just hardware; it’s a complete HPC environment tailored to Indian AI development,” says a government official familiar with the program. Whether GNANI.ai or other domestic players will tap into the package through subsidized access remains a central question in the broader IndiaAI ecosystem.

Why IndiaAI now?

The IndiaAI Mission carries an outlay of over ₹10,300 crore and is designed to seed a robust, domestic AI ecosystem. The programme targets the deployment of more than 10,000 GPUs initially for use by startups, researchers, and academic institutions. In practice, this policy aims to level the playing field: instead of most cutting-edge AI development happening offshore, Indian firms can train models on locally sourced data and with Indian-language and cultural nuances in mind.

Gnani.ai’s role as the recipient highlights a broader push to back indigenous foundational models. The government’s strategy envisions a pipeline of models tailored to India’s datasets, languages, and contexts—reducing exposure to foreign models that may not capture India’s diversity. Subsidized compute is at the heart of this strategy, lowering the barrier to entry for researchers and startups across the country.

E2E Networks: a domestic AI infrastructure pillar

E2E Networks is racing to position itself as the go-to AI-focused hyperscale cloud platform within India. Backed by Larsen & Toubro, the company has been expanding its footprint in the domestic cloud scene and recently acquired assets from Jarvis Labs AI, a Coimbatore-based GPU cloud services startup. The timing of the ₹177 crore government order is being read as a strong signal of market validation—investors cheered the move, and E2E’s stock price reflected that optimism.

This isn’t just about a single contract; it’s part of a larger narrative where Indian cloud providers build the compute backbone for national AI ambitions. The Jarvis Labs acquisition, in particular, signals E2E’s intent to deepen its AI and ML offerings, creating a more cohesive ecosystem for training and inference within India’s borders.

What could this mean for India’s AI landscape?

• Sovereignty and security: Domestic compute means more control over data governance and model training pipelines.
• Language and culture: Foundational models trained on Indian languages and contexts could deliver more relevant results for local users and businesses.
• Startup acceleration: Subsidized access can lower the bar for early-stage research and product development.
• Global competitiveness: A credible domestic HPC stack helps Indian startups compete on a global stage without outsourcing critical compute.

However, the path isn’t without hurdles. Building and maintaining a national-scale AI compute layer requires ongoing funding, robust data governance, and a talent pool capable of building and scaling such infrastructure. The MeitY plan will need to balance resource allocation with the needs of a diverse community of researchers, startups, and institutions that have different tolerance for risk and cost.

Market response and next steps

News of the contract caused a notable reaction in the Indian tech-finance space. Investors are watching closely to gauge whether India’s national AI infrastructure can translate into tangible products and services that reach millions of users. If the program sustains momentum, India could emerge as a serious alternative to traditional AI hubs, offering models and tools that are optimized for its own data regimes.

As the IndiaAI Mission unfolds, the focus will be on how quickly and effectively the domestic ecosystem can translate subsidized compute into usable AI systems, from industrial automation to healthcare or education. The coming quarters should reveal how Gnani.ai and other players leverage this GPU capacity to push forward indigenous AI capabilities.

Conclusion

The ₹177 crore deal for E2E Networks marks more than a procurement win—it’s a signal that India intends to build its own AI future. By empowering a domestic cloud player to supply a major Indian startup with state-of-the-art computing resources, the IndiaAI Mission is actively cultivating a self-reliant AI ecosystem. The goal isn’t just to copy what’s happening elsewhere; it’s to craft AI that’s attuned to India’s data, languages, and needs. If successful, this initiative could redefine how large-scale AI is built in India, paving the way for technology that’s designed in India, for India.