Indian Tech Giants Deploy NVIDIA Nemotron for Enterprise AI Agents

Ted Hisokawa Feb 18, 2026 00:55

Infosys, Wipro, Tech Mahindra and Persistent build agentic AI platforms using NVIDIA's open models, targeting $500B Indian tech market by 2030.

India's largest IT services firms are betting big on NVIDIA's Nemotron open models to power a new generation of enterprise AI agents, with deployments already showing measurable results in healthcare, telecommunications, and software development.

Infosys, Wipro, Tech Mahindra, and Persistent Systems unveiled their NVIDIA-powered agentic AI platforms at the India AI Impact Summit, signaling a major push into autonomous business operations. The timing aligns with India's tech sector targeting $500 billion in revenue by 2030—double its 2023 figure of approximately $250 billion, according to IBEF data.

Wipro's Call Center Results Stand Out

Wipro's WEGA platform, deployed for a major U.S. healthcare insurer, delivered the most concrete performance metrics. AI agents now handle 42% of inbound calls with sub-200-millisecond latency, managing 900 concurrent calls and 164 requests per second. For a sector drowning in seasonal enrollment surges and compliance requirements, those numbers translate to real cost savings.

The system runs on NVIDIA NIM microservices and NeMo Guardrails—critical for healthcare's regulatory environment where a chatbot hallucination isn't just embarrassing, it's potentially illegal.

Tech Mahindra Targets Telecom's $1.5T Market

Tech Mahindra's approach focuses on network operations, building a large telco model that ranks repair recommendations by historical success rates. The goal: first-visit fixes and fewer escalations. They're using Nemotron embedding models for semantic search across network telemetry, with a reranking model to prioritize decisions.

A large telecom provider has already adopted the platform. In an industry generating over $1.5 trillion annually, even marginal uptime improvements compound quickly.

Infosys Goes Small for Coding

Infosys took a different path, developing a 2.5-billion-parameter coding model—deliberately compact compared to frontier models. Built with NVIDIA's NeMo framework, it matches larger models on benchmarks like MBPP and BFCL while running on standard enterprise hardware or desktops.

The model passed Stanford AIR-Bench and Meta's CyberSecEval for secure coding validation, addressing enterprise concerns about AI-generated code vulnerabilities.

Persistent Tackles Drug Discovery

Persistent Systems' GenMoIVS solution applies the same agentic architecture to molecular discovery, using NVIDIA BioNeMo to simulate compound behavior before wet lab testing. The pitch: de-risk early drug development and shorten timelines in an industry where failed candidates cost billions.

Why Open Models Matter Here

These deployments build on NVIDIA's Nemotron 3 family, announced in December 2025 with a hybrid mixture-of-experts architecture supporting up to 1 million token context lengths. The Nano model—available now with Super and Ultra versions coming in early 2026—offers pricing starting at $0.06 per million input tokens on AWS Bedrock.

NVIDIA's decision to open-source model weights and training data gives these Indian integrators the customization flexibility enterprise clients demand. ServiceNow, Bosch, CrowdStrike, and Palantir have already signed on as early adopters elsewhere.

For investors tracking NVIDIA's enterprise AI momentum, India's systems integrators represent a significant distribution channel. With 38,000 GPUs secured in September and four major IT firms building production platforms, the Nemotron ecosystem is moving from benchmarks to billable deployments.

• nvidia
• enterprise ai
• nemotron
• india tech
• agentic ai