Key Takeaways
- Sarvam AI has unveiled Chanakya, a new applied-AI vertical after more than 12 months of quiet development. The focus was on “problems of national consequence and complex enterprises.”
- Chanakya offers air‑gapped, on‑prem AI deployments with multimodal data support. It also provides production‑grade agentic workflows for high‑stakes government and regulated enterprise use.
- The systems are designed for dual use across strategic sectors such as defence and national security. They also serve large enterprises that cannot rely on public cloud infrastructure.
- Sarvam AI is reportedly in early talks to raise around 250–300 million dollars. The goal is to scale its sovereign AI stack and Chanakya deployments.
Quick Recap
Indian AI startup Sarvam AI has announced Chanakya, a new applied-AI vertical built over the past year. The goal is to address “problems of national consequence and complex enterprises.” In an official post on X (formerly Twitter), the company said it is now scaling this effort into a dedicated line of mission‑critical, high‑security AI services for governments and large institutions.
Building an AI Stack for High-Security Missions
Sarvam AI describes Chanakya as an applied AI service that combines its full‑stack models, tooling, and deployment infrastructure into offerings tailored for environments “where failure isn’t an option.” The stack supports on‑premise deployments in air‑gapped data centres. Furthermore, it enables multimodal data ingestion across text and images, and production‑grade agentic workflows designed to run autonomously under strict reliability and security constraints.
Under the hood, Chanakya can leverage Sarvam’s homegrown models, including Sarvam‑30B and Sarvam‑105B. These support 22 Indian languages and power products such as Sarvam Vision (OCR and multimodal), Sarvam Dub (translation and dubbing), and the Indus beta app. According to recent reports, Sarvam AI is also in early discussions to raise approximately 250–300 million dollars from investors such as Nvidia, HCLTech, and Accel. This capital would likely underpin Chanakya’s rollout across government and enterprise accounts.
Why This Move Matters Now?
Chanakya lands at a moment when India is pushing for “sovereign AI” infrastructure and stricter data‑governance regimes for critical sectors. Many government departments, defence agencies, and regulated enterprises cannot move sensitive workloads to global public clouds. This creates demand for on‑prem, air‑gapped AI stacks that still match frontier capabilities.
Sarvam AI positions itself as “sovereign by design,” and Chanakya extends that thesis from foundational models to full production systems for national‑scale problems. The vertical also responds to a broader shift in the AI market, where vendors are racing to offer agentic, workflow‑oriented platforms. These platforms embed AI directly into mission‑critical operations rather than just chat interfaces.
Competitive Landscape and Capability Comparison
For Chanakya, two relevant India‑focused, sovereign or secure‑AI peers are Krutrim’s enterprise/sovereign AI stack and Haptik’s Generative AI for Enterprises platform. Both also target regulated or large‑scale deployments with local language support. Public, precise metric-by-metric data is limited, but typical positioning can be outlined as follows based on available disclosures and market reports.
Applied AI Stack Comparison: Chanakya vs Peers
| Feature/Metric | Chanakya (Sarvam AI) | Krutrim Enterprise/Sovereign AI* | Haptik GenAI for Enterprises* |
| Context Window | Optimised for long‑form, high‑stakes workflows; exact token limits not disclosed | Large‑context models for enterprise search and assistants; details sparse | Tuned for customer support and workflows; uses extended context via orchestration |
| Pricing per 1M Tokens | Enterprise and government contracts; per‑deployment / usage‑based, not public | Custom enterprise pricing; likely similar contract‑driven structure | SaaS or usage‑based pricing layered over APIs; rates not publicly broken out per 1M tokens |
| Multimodal Support | Yes – supports text and images via Sarvam Vision and related tools | Emerging multimodal capabilities, focused on Indian languages and voice | Primarily text and voice for CX; limited public info on full image support |
| Agentic Capabilities | Production‑grade agentic workflows for air‑gapped, mission‑critical environments | Enterprise automation and copilots; agentic behaviour around business apps | Workflow and CX agents across chat, support, and CRM integrations |
While Chanakya appears to lead on air‑gapped, defence‑ and government‑grade deployments with deep multimodal and agentic support, Krutrim’s stack may be more broadly positioned across India‑centric language and cloud environments. Haptik remains strong for customer‑experience‑focused agents embedded in existing SaaS tools. For high‑stakes missions where regulatory risk and uptime trump cost transparency, Chanakya’s design choices give it a notable edge. In contrast, CX‑heavy or cloud‑native enterprises might still favour the more established SaaS‑style offerings from competitors.
Sci-Tech Today’s Takeaway
I think Chanakya is a big deal because it pushes India’s AI story beyond generic chatbots into the messy, high‑stakes world of defence, governance, and regulated enterprise infrastructure. In my experience, when a startup commits to air‑gapped deployments, multimodal workflows, and agentic systems for “problems of national consequence,” it is signalling a long‑term bet on sovereign, infrastructure‑level AI rather than just app‑layer experiments. I generally see this as bullish for serious enterprise and government adoption: it gives risk‑averse institutions something they can actually deploy under their own security models. At the same time, it also forces competitors to raise the bar on reliability and control.
Sources
