Yotta Built India’s First Sovereign AI Infrastructure With Shakti Cloud

Yotta is transforming India’s AI landscape with cutting-edge cloud and data center solutions, accelerated by NVIDIA’s AI platform. From deploying high-density data centers to democratizing AI access with a pay-per-use model, Yotta is making compute for AI affordable and accessible for businesses of all sizes.

India has 1.4 billion people and over 300 distinct languages and dialects, but the country lacked AI models trained on local languages and cultural contexts. Yotta recognized the necessity of developing sovereign AI models to cater to the diverse linguistic and cultural landscape of the country.

To differentiate itself from global competition, Yotta needed to offer unique services tailored to the Indian market. The goal was to provide affordable compute infrastructure and the tools needed for Indian startups and enterprises to build their own AI applications. Ultimately, Yotta wanted to democratize access to the full-stack AI ecosystem by offering cost-effective solutions, including flexible payment models like accepting equity instead of cash.

By leveraging NVIDIA’s advanced, comprehensive stack for AI Infrastructure, Yotta could compete aggressively with existing vendors, especially on localized and sovereign offerings. The company identified that NVIDIA H100 Tensor Core GPUs and NVIDIA A100 Tensor Core GPUs, along with NVIDIA’s global AI leadership and full-stack ecosystem—including hardware, software, and software development kits—could deliver the scale and performance to bring Shakti Cloud to life.

Overall, Shakti Cloud has over 16,000 H100 GPUs, with more on the way, including NVIDIA B200 Tensor Core GPUs. The integration of NVIDIA reference architecture has translated to high-speed interconnects such as NVIDIA NVLink™ and NVIDIA Quantum-X800 InfiniBand to effectively eliminate critical data transfer bottlenecks in Yotta’s AI and HPC workloads. NVIDIA’s networking technologies have substantially reduced training times and improved inference latency, enhancing overall pipeline efficiency.

Yotta also takes advantage of NVIDIA’s extensive software stack. NVIDIA AI Enterprise has further accelerated Yotta’s platform development for its customers by providing production-grade, optimized containers and pretrained models, serving as a robust foundation for diverse AI deployments.

Furthermore, NVIDIA Cloud Functions (NVCF), a serverless API to deploy and manage AI workloads, has enabled scalable, low-latency inference services, allowing Yotta’s customers to efficiently scale their AI-driven applications. Collectively, these capabilities have been instrumental in delivering differentiated, high-performance AI infrastructure.

Yotta is proud to be India’s first (and among only five globally) Reference Platform NVIDIA Cloud Partner (NCP). By integrating NVIDIA’s advanced AI technologies into Yotta’s services, it provides clients with instant access to state-of-the-art tools and resources through its Rudra platform. The self-service portal and advanced orchestration layer within Shakti Cloud offer a seamless user experience, allowing customers to focus on their AI projects without the burden of infrastructure management. This enhances both customer satisfaction and operational efficiency.

Yotta’s infrastructure and access to NVIDIA AI Enterprise software have significantly accelerated the development of large language models (LLMs) and small language models (SLMs) in India able to serve India’s massive population and variety of local language and dialects, as well as helped position the country as a global AI hub.

One of Yotta’s customers, Sarvam AI, builds LLMs to cater to India’s distinct population needs. Sarvam has LLMs in 10 different languages, and voice-first agentic AI solutions for call centers and WhatsApp integration. Sarvam has realized a 20%–100% improvement in throughput compared to open-source options alone for serving AI models. It has also reduced the time needed to manage Kubernetes clusters and manage auto-scaling deployments for serverless deployments.

Through Shakti Cloud, Yotta sees generative AI having a profound impact on various industries in India, especially in healthcare for personalized medicine, and in legal tech for document analysis. The company believes that these types of applications will drive innovation and efficiency in sectors critical to India’s economy.

Yotta continues to expand its sovereign AI cloud footprint, with plans to scale to over 32,768 GPUs for its Telangana AI infrastructure by the end of 2025. The company also plans to enhance its software offerings with NIM™ microservices to build a robust ecosystem that fosters innovation within the Indian AI landscape.

Yotta’s ultimate objective is to offer AI infrastructure and solutions at an affordable price point and enable India’s mission of being the AI infrastructure and innovation hub of the world.