Yotta needed to deliver affordable, sovereign AI infrastructure inside India so startups and enterprises could build local-language models without depending on overseas compute platforms that lacked India-specific capacity and economics.
Reported impact
Yotta created India's first adapted AI infrastructure platform, scaled Shakti Cloud to more than 16,000 H100 GPUs, improved training and inference efficiency through NVIDIA networking, and enabled customers such as Sarvam AI to improve model-serving throughput by 20 percent to 100 percent.
Yotta built Shakti Cloud as a localized AI infrastructure platform on NVIDIA's full-stack ecosystem so Indian customers can train and serve models on domestic infrastructure. The platform combines H100 and A100 GPU capacity today, planned B200 expansion, NVLink and Quantum-X800 InfiniBand reference networking, AI Enterprise software, and NVIDIA Cloud Functions for serverless inference. Around that stack, Yotta exposes a self-service orchestration layer and pay-per-use commercial model that lets customers consume GPU clusters, pretrained software, and scalable inference services without building their own AI platform. That architecture turns sovereign compute, networking, model-serving, and operational tooling into one cloud product that supports Indian-language LLMs and SLMs while giving Yotta room to scale from its current footprint toward a much larger national AI supercloud.
Customer training and inference jobs are provisioned through Yotta's orchestration layer onto Shakti Cloud GPU clusters, where InfiniBand and NVLink-backed infrastructure move data across accelerated nodes. AI Enterprise provides optimized software images, and Cloud Functions exposes serverless inference endpoints for applications built on the same platform.
Furthermore, NVIDIA Cloud Functions (NVCF), a serverless API to deploy and manage AI workloads, has enabled scalable, low-latency inference services.
India has 1.4 billion people and over 300 distinct languages and dialects.
Overall, Shakti Cloud has over 16,000 H100 GPUs, with more on the way, including NVIDIA B200 Tensor Core GPUs.
Sarvam has realized a 20%-100% improvement in throughput compared to open-source options alone for serving AI models.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph NVIDIA Sample Sep 15, 2026 Lambda Lambda needed to raise token throughput and improve AI-cloud economics, but fixed rack and facility power budgets were leaving GPU capacity stranded and limiting how many nodes could run safely at once. +17% Hybrid Training Throughput Dynamic Power Software · NVIDIA DSX AI Factory Platform · NVIDIA DSX MaxLPS NVIDIA Sample Sep 1, 2026 Onetera Onetera needed to help more than 90 municipalities turn fragmented regulations, zoning maps, permit rules, and legacy documents into fast resident-facing services without imposing heavy IT overhead or expensive manual… 1.7x increase in data throughput GPT OSS 20B · Government Intelligence Agent · LangGraph 