Mukesh Ambani Expands Jio Into AI-Powered Calls, Apps, Homes

At the AGM on June 19, Mukesh Ambani said Reliance Intelligence, launched last year, is moving into an "execution" phase. He framed the push in explicitly nationalistic terms: "India should not be a mere consumer of AI created elsewhere. It must become a creator, adopter, and global leader in AI," he said, per the Chairman's Statement and TechCrunch's reporting. Concretely, Reliance is building a sovereign AI compute backbone in Jamnagar, powered by its Kutch solar platform, with the first 120 MW due by end-2026 running an initial NVIDIA GB300 GPU fleet that Reliance says is already equivalent to more than 75,000 H100 GPUs on an inference basis, scaling past 200,000 H100-equivalents once the site is fully operational. TechCrunch and Indian Express report the consumer layer sitting on top: an AI-powered MyJio, which Reliance says now has more than 600 million users, handling requests like eSIM activation and roaming plans conversationally, plus Jio TeleFrame, a home display for AI agents. Times of India reports, citing PTI, that Jio Platforms filed draft IPO papers the same day, seeking to raise about $4 billion at a valuation of roughly $137 billion.

The multilingual requirement is a tougher production-ML target than the demos suggest. Reliance says its wider AI stack, including JioBharatIQ, AI Vyapar, JioHealthIQ, JioLearnIQ, and JioKrishiIQ, must work across 22 Indian languages, and the Chairman's Statement says the call agent itself works "in every Indian language" on live calls, not just in text chat. Real-time speech recognition, speaker diarization, summarization, and task execution, stacked across dozens of languages at telecom scale, is a substantially harder deployment target than a single-language chatbot, and it will be a useful public test of how far carrier-grade multilingual voice AI has actually progressed once it ships.

TechCrunch's reporting flags a detail worth tracking: Reliance said its AI features operate only with user consent but did not clarify whether call and app data will be used to train its models or shared with partners such as Google and Meta, whose free Gemini access and joint LLaMA venture Reliance is already building on. That data-governance gap looks more consequential than any individual product feature, given the volume of conversational data involved.

The widely cited Rs 10 lakh crore (about $110 billion), seven-year AI infrastructure commitment did not originate at this AGM. Ambani first announced it in February 2026, as DD News and The Logical Indian both report, and the AGM materials add evidence that the commitment is converting into commissioned capacity on a specific timeline, which is the more verifiable signal for practitioners tracking whether announced AI capex actually gets built. Moneycontrol and Firstpost report further pieces of the buildout: Jio Platforms managing director Akash Ambani outlined the 22-language roadmap and said Jio is separately evaluating a sovereign low-earth-orbit satellite constellation, while an existing internal platform, JioBrain, already handles network planning and predictive maintenance.

• •Whether Reliance publishes transcription and diarization accuracy data for Jio Call Agent once it launches for Jio's 500-million-plus base later this year.
• •Whether the 120 MW Jamnagar milestone and GB300 deployment land on the stated end-2026 timeline, and whether model architecture or training-data sourcing gets disclosed.
• •How Reliance answers the data-use and consent questions TechCrunch raised, given the scale of conversational and app data involved.