Reliance Intelligence operationalises 120MW AI backbone in Jamnagar

At Reliance Industries' 49th annual general meeting, Jio Chairman Akash Ambani announced that Reliance Intelligence will commission the first 120 MW of AI compute capacity at a Jamnagar data centre by end 2026. The facility is powered entirely by renewable energy from Reliance's Kutch solar platform, per CNBC-TV18 and The Hindu BusinessLine.

Reliance is deploying an initial fleet of NVIDIA GB300 GPUs. Ambani stated the capacity delivers compute equivalent to more than 75,000 H100 GPUs on an AI-inference basis and described a path to over 200,000 H100-equivalent GPUs as the full 120 MW becomes operational (CNBC-TV18, Economic Times). Practitioners should treat vendor-equivalence figures - GB300-to-H100 conversion based on inference throughput - as capacity proxies rather than direct training-performance benchmarks, as architecture, power draw, and workload mix differ significantly between GPU generations.

The Jamnagar build is part of Reliance's previously announced $110 billion AI infrastructure commitment (Reuters, TechCrunch). Large renewable-powered AI campuses reduce exposure to grid price volatility when sited near captive generation. Reliance's stated goal is to serve 22 Indian languages through applications including JioBharatIQ, JioHealthIQ, JioLearnIQ, and JioKrishiIQ. Akash Ambani described the project as India's "sovereign AI backbone." At the same AGM, Jio filed its IPO draft with SEBI - a separate announcement that underscores the broader scale of Reliance's public market ambitions.

Three operational indicators will determine whether the announced capacity converts to usable compute: rack-level cooling and integration choices as Reliance moves GB300 installs toward full 120 MW operation; power availability and grid-interconnection performance as large variable loads come online alongside captive solar; and ecosystem outcomes - developer access models, inference pricing, and enterprise or hyperscaler partnerships that determine whether the capacity reaches Indian ML practitioners at scale. Reliance has not published benchmarked per-model performance data or commercial pricing details as of the AGM.