Podstack is India's sovereign AI Infrastructure Operating System. 

We deliver fractional GPU compute, MLOps tooling, and per-token managed inference — all inside India's data borders. 

From a self-serve developer signup at ₹45/hour to a DPDPA-architected enterprise reserved tier for BFSI and healthcare, Podstack is built for the buyers who are structurally underserved by hyperscalers and structurally under-protected on commodity neoclouds.

At the core is PodVirt — a Rust-based, software-defined GPU virtualisation layer that runs across NVIDIA H100, L40S, A100, L4, and AMD MI300X without vendor SDK lock-in. It is the wedge that lets a regulated Indian buyer transact under one Podstack contract for sovereign managed inference, while compute is federated across our partner data centres.

Who we serve: gen-AI startups, ML research teams, mid-cap BFSI, mid-tier healthcare and life sciences, MoD-aligned workloads — anyone who needs performance, sovereignty, and predictable INR pricing.

What we serve: Fractional GPU (PodVirt), Serverless GPU, Baremetal, Pods, VMs, Object Storage, NFS Volumes, Serverless Jupyter Notebooks, Model Registry, One-Click Inference, MLOps Pipelines, and Managed Inference at per-token pricing.

Why we exist: India has a structural compute gap. Hyperscalers cannot economically serve mid-enterprise procurement. Commodity neoclouds cannot provide the governance plane a regulated CIO can put through compliance. Foundation-model providers will not localise their inference plane to Indian DPDPA architecture. 

Podstack is the fourth, deliberately constructed position.Get started: visit podstack.ai. Self-serve signup is open. KYC unlocks ₹X in free credits.

HQ: India   
·   Live in production   
·   Hiring: ML systems engineers, SREs, BFSI account executives.

PodStack

Org chart

Teams

Offices