NVIDIA and Reliance Jio have announced a strategic partnership to deploy a large-scale AI GPU infrastructure across Jio's Indian data centres, focused on training Indian language large language models and supporting AI inference workloads for enterprise and consumer applications. The cluster will use NVIDIA Hopper and Blackwell GPU generations, making it among the most capable AI compute resources outside government-sponsored hyperscalers in India.
Semicon Hunt -> partnership -> NVIDIA / Reliance Jio
2026-09-01
NVIDIA and Reliance Jio have formalised a strategic technology partnership to build and operate a large-scale GPU-based AI compute cluster across multiple Jio data centre locations in India. The collaboration combines NVIDIA's flagship AI accelerator hardware, specifically the Hopper H100 and the newer Blackwell B200 GPU generations, with Jio's national data centre infrastructure and its extensive telecommunications backbone. The partnership is positioned as the foundation for training and deploying AI models tailored specifically to Indian languages, dialects, and use cases, addressing a gap in global AI development where models have historically underperformed on low-resource Indian language tasks.
The cluster will aggregate GPU resources across multiple Jio data centre facilities, creating a compute pool that can be accessed by Jio's enterprise clients, government departments, and Jio's own consumer AI service teams. NVIDIA is supplying both H100 SXM5 and B200 GPU modules, configured in DGX SuperPOD rack systems with NVIDIA's NVLink high-bandwidth interconnect fabric. The NVLink topology allows model training to span thousands of GPUs as a single logical compute unit, which is necessary for training large language models and multimodal foundation models at the parameter scales required for human-level performance across diverse Indian language domains. NVIDIA is also deploying its networking infrastructure including InfiniBand switches to minimise GPU idle time during distributed training runs.
Reliance Jio serves over 450 million mobile subscribers and has deep datasets of voice, text, and transactional data in Hindi, Tamil, Telugu, Bengali, Marathi, and dozens of other Indian languages. The GPU cluster enables Jio to train large language models on this proprietary data corpus, creating models that understand colloquial usage, regional vocabulary, and code-switching between languages and English. Jio's AI products team plans to deploy these models in Jio's consumer assistant applications, enterprise contact centre automation services, and connectivity management platforms. NVIDIA's AI Enterprise software stack will be deployed alongside the hardware, giving Jio's data science teams access to optimised training frameworks, inference servers, and monitoring tools.
The NVIDIA-Jio partnership is significant for India's AI ecosystem because it creates a large private compute resource that will be partially available to third-party developers and researchers through cloud access models, reducing the barrier for Indian AI startups that cannot afford to build dedicated GPU clusters. NVIDIA has indicated that India is now one of its fastest-growing markets for data centre GPU systems and that the Jio partnership is part of a broader effort to establish India as a significant node in the global AI training and inference infrastructure map. For Reliance, the investment positions Jio as more than a telecom operator, establishing it as a full AI infrastructure provider competing with Amazon Web Services, Microsoft Azure, and Google Cloud in the Indian enterprise market.
The form has been successfully submitted.
We will review your software soon!
See you soon.
Please enter your details and we will