Shakti Cloud
6/10Shakti Cloud is Yotta Data Services' sovereign GPU cloud in India. It offers H100, L40S and B200 virtual machines, bare metal, clusters and serverless inference for AI teams.
Data last updated November 2026 · Our methodology
Overview
Shakti Cloud is the AI infrastructure platform of Yotta Data Services, a Hiranandani group company. It runs on NVIDIA H100 and L40S GPUs, with B200 bare metal listed, in Yotta's data centres in Navi Mumbai and Greater Noida. Products include AI Workspace GPU virtual machines, dedicated bare-metal GPU servers, SLURM and Kubernetes training clusters, serverless GPUs, OpenAI-compatible AI Endpoints for hosted models, fine-tuning, and AI Lab environments for universities and research groups. It suits Indian enterprises, government bodies, BFSI firms and AI startups that need data kept in India under the DPDP Act and want to be billed in rupees. Yotta is an NVIDIA Cloud Partner, and the LLM builder Sarvam AI trains its models on the platform. Pricing is published in INR. On-demand VMs are billed hourly, for example 356 rupees per hour for one H100. Dedicated VMs are billed monthly, and bare metal and clusters are priced per GPU-hour, with commitment terms of up to five years. InfiniBand and a Kubernetes control plane are charged as extras. The platform includes NVIDIA AI Enterprise and NIM, and it connects to Azure OpenAI and Azure ML. Its main limitation is scope: this is a GPU and AI cloud, not a general-purpose cloud. It has no web hosting, managed databases or CDN. Its on-demand H100 rate is also higher than some Indian rivals such as Cyfuture AI and JarvisLabs.
Our Rating
Hosting & Infrastructure6.1/10
AI Software5.8/10
Who Shakti Cloud is best for
Good fit
It offers large H100 clusters, multi-year commitments and Indian data residency, aimed at enterprise AI programmes.
It is an IndiaAI Mission empanelled provider with sovereign data centres, and coverage places it with government customers.
AI Lab is a dedicated managed product for universities and research groups.
Its data centres are described as RBI-certified, and a comparison article places it with BFSI customers.
Less suited
It provides OpenAI-compatible endpoints and API provisioning, but only for AI workloads.
It offers on-demand hourly GPUs and serverless endpoints, and Sarvam AI builds on it, but rates are not the cheapest.
Pros & Cons
Pros
- It publishes INR prices for VMs, bare metal and clusters, so Indian buyers avoid exchange-rate risk and can estimate costs without a sales call.
- Data centres in Navi Mumbai and Greater Noida keep data in India for DPDP Act compliance.
- It covers the full range from hourly single-GPU VMs to 256-plus-GPU H100 clusters over InfiniBand, as one of NVIDIA's Cloud Partners.
- Unlimited ingress and egress are included on VMs, and data transfer is included on bare metal and clusters.
- It bundles NVIDIA AI Enterprise and NIM, plus OpenAI-compatible serverless endpoints for hosted open models.
Cons
- The on-demand H100 VM, at 356 rupees per hour, costs more than Cyfuture AI (329) and JarvisLabs (about 330), according to a comparison article.
- It is limited to GPU and AI workloads. It has no general web hosting, managed databases, object storage pricing or CDN.
- InfiniBand and the Kubernetes control plane cost extra, and term discounts are not published.
- Independent user reviews are scarce, and sources disagree on whether access is self-serve or quote-based.
Key Features
Shakti Cloud Pricing
AI Lab (1x H100 80GB workstation)
- 8x H100 at 1,580,000 rupees per month
- L40S workstations from 81,000 rupees per month
- H100 slices from 30,000 rupees per month for 10GB
AI Workspace VM Dedicated (1x H100)
- 500 GB root disk
- Unlimited ingress and egress
- Terms of 1, 3, 6 or 12 months
- 1x L40S at 80,221 rupees per month
- 4x H100 at 832,200 rupees per month
AI Workspace VM On-Demand (1x H100)
- 100 GB root disk
- 356 rupees per hour
- Unlimited ingress and egress
- 1x L40S at 137 rupees per hour
Bare Metal (8x HGX H100)
- 428 rupees per GPU-hour
- Terms from monthly up to 5 years
- 8x HGX B200 at 473 rupees per GPU-hour
- BlueField-3 DPU and data transfer included
Cluster (SLURM, 8x HGX H100 SXM5)
- Data transfer included
- 386 rupees per GPU-hour
- Terms from monthly up to 2 years
- 4x L40S at 217 rupees per GPU-hour
Frequently Asked Questions
Yes. The pricing page lists INR rates, for example 356 rupees per hour for an on-demand H100 VM and 428 rupees per GPU-hour for 8x H100 bare metal. Large commitments may still be negotiated.
In Yotta's own data centres in Navi Mumbai and Greater Noida, India. The vendor states it follows the DPDP Act and Indian data residency rules.
NVIDIA H100 and L40S across VMs, bare metal and clusters. HGX B200 bare metal is priced, and B300 is listed without a price.
Yes. Clusters can run SLURM or managed Kubernetes. The Kubernetes control plane is charged monthly, from 12,951 rupees for non-HA and 38,853 rupees for HA.
The pricing page lists unlimited ingress and egress for VMs and serverless, and includes data transfer for bare metal and clusters.
The site says its data centres are RBI-, ISO- and SOC-certified. Yotta's separate Yntraa cloud is MeitY-empanelled.
Similar Tools
See all alternatives
Amazon's public cloud platform, with over 200 pay-as-you-go services for compute, storage, databases, networking and AI, used by startups through large enterprises and governments.
Google's public cloud platform providing pay-as-you-go virtual machines, Kubernetes, serverless, databases, storage, networking, data analytics and AI services for businesses of every size.
Northflank is a Kubernetes-based developer platform for deploying services, jobs, databases and GPU workloads on its managed cloud or on customers' own cloud accounts.