IBM and Yotta open sovereign agentic AI platform for Indian firms

BENGALURU: IBM and Yotta Data Services on 29 September 2026 announced general availability of a Sovereign Agentic AI Platform for Indian organisations. The stack pairs IBM watsonx Orchestrate with Yotta's Shakti Cloud so enterprises can deploy and govern AI agents while keeping data, inference and operations inside India.

IBM describes watsonx Orchestrate as the agentic control plane for securing and managing agents as use spreads across a company. Yotta's Shakti Cloud supplies compute, GPU capacity, networking, security and operations for those workloads. Shakti Studio, Yotta's model platform, is positioned for exploring, fine-tuning, deploying and running open-source and proprietary models on the same sovereign footing.

The platform is available through Yotta's Panvel and Greater Noida cloud regions. Suggested early workflows include security operations, document processing and HR automation, though IBM and Yotta framed the offer as broader once governance controls are in place.

Why sovereignty is the pitch

Sandip Patel, Managing Director of IBM India and South Asia, said evolving Indian rules make digital sovereignty and open-source models a defining requirement as firms move from pilots to production. He argued the priority is no longer only what AI can do, but how securely and compliantly it can be deployed, with critical data, inference and governance anchored in India.

Sunil Gupta, Co-Founder, MD and CEO of Yotta Data Services, said organisations need control not only over data but over infrastructure, compute, models, inference and operations. He cast Shakti Cloud and Shakti Studio as that sovereign foundation, with IBM adding enterprise agentic controls on top.

Yotta notes that its Yntraa hyperscale cloud is MeitY-empanelled and that Shakti Cloud includes AI labs, workspaces, inference services, NVIDIA NIM access and GPU Kubernetes clusters. Yotta also cites NVIDIA Cloud Partner status in the APAC region. Those claims matter for public-sector and regulated buyers who already shortlist MeitY-aligned infrastructure.

What buyers still need to confirm

Pricing is not listed as a public rate card; enterprise deals will turn on capacity, GPU class, support and compliance scope. Teams evaluating the platform should map which agent workflows truly need India-resident inference, which models they will bring, and how watsonx Orchestrate ties into existing identity and audit systems.

General availability removes the "pilot only" tag from the IBM-Yotta pairing for Indian buyers who have been waiting for a production path. It does not replace a formal security review. Boards that care about data residency now have a named stack, two NCR and Mumbai-adjacent regions, and quoted executives putting sovereignty at the centre of the sale.

Indian banks, insurers, hospitals and government agencies have spent two years asking vendors where inference runs and who can see prompts. A jointly branded GA release from IBM and a MeitY-empanelled Indian cloud operator answers that question more directly than a slide deck about "hybrid options". Panvel and Greater Noida regions also give buyers a west-and-north footprint instead of a single-city dependency.

Integration work remains. Identity providers, existing SOC tools and model approval boards still need to plug into watsonx Orchestrate. Firms should demand a reference architecture for one workflow, such as document processing with India-only logs, before signing a multi-year GPU commit. GA means the product is sellable; it does not mean every regulated workload is automatically cleared.