Reliance Intelligence Ditches NVIDIA GPUs for Google TPUs

The future of the Indian AI ecosystem looks promising with Reliance entering the scene. The company that reshaped the telecom landscape with Jio now aims to replicate its success in AI through Reliance Intelligence, launched in partnership with Meta and Google. 

The timing couldn’t be more perfect, as reports have emerged that OpenAI is currently seeking local partners to develop a 1 gigawatt data centre in India, which is its second biggest market after the US. Previously, it was reported that OpenAI was also looking to partner with Reliance. 

However, it appears that the deal didn’t go through, and Reliance partnered with Meta instead, which had also been in the discussions.

At the same time, Reliance is also building a data centre in Jamnagar with a total capacity of three gigawatts. It will surpass existing facilities, which typically operate at less than one gigawatt.

During the NVIDIA Summit in Mumbai last year, CEO Jensen Huang had a fireside chat with Mukesh Ambani, where he announced an upgraded deployment of Blackwell AI processors for a 1-gigawatt data centre in Jamnagar. 

This was a follow-up to the September 8, 2023 announcement, when Reliance Industries’ telecom arm, Jio Platforms, entered a partnership with NVIDIA to develop India’s own foundational LLM for diverse languages. 

As part of the agreement, NVIDIA committed to supplying its advanced GH200 Grace Hopper Superchip and DGX Cloud supercomputing services. 

There had been no updates on the partnership until Google announced a fresh collaboration with Reliance. “Google and Reliance are partnering to transform all of Reliance’s businesses with AI—from energy and retail to telecom and financial services. To enable this adoption, we are creating a dedicated cloud region for Reliance, bringing Google Cloud’s AI and compute, powered by Reliance’s clean energy and connected by Jio’s advanced network,” said Sundar Pichai, CEO of Google, at Reliance AGM 2025 in a pre-recorded video.

“Reliance announced a partnership with NVIDIA about a year back, with Jensen & Mukesh having a chat. They seem to have dropped that process with NVIDIA a year later and aligned with Google for infrastructure and Meta for LLMs,” A S Rajgopal, CEO and MD at NxtGen Cloud Technologies, told AIM.

AIM reached out to Reliance for an update on the deal, but did not receive a response.

Rajgopal said that “Reliance Intelligence will essentially serve as a front-end to open-source models from Meta, running on Google Tensor Processing Units (TPUs),” adding that Stargate, in contrast, is focused on deploying “much larger and more versatile GPU infrastructure.” 

Reliance is also using AMD GPUs. The company has already deployed a significant proof of concept (POC) with Jio and is working on future deployment plans that are yet to be announced. 

Rajgopal added that open-source models are rapidly catching up with systems like ChatGPT in performance and capabilities, and since these models are publicly available, Reliance does not have exclusive access.

He further added that Meta & Google have done a lot of work on Indian languages, and both of them have large data pertaining to Indians. The new alliance between Reliance, Google and Meta could benefit Indian citizens with services that are contextually Indian. 

Can Reliance Challenge OpenAI?

For Reliance to succeed in the consumer segment, it needs to challenge OpenAI, which recently introduced ChatGPT Go in India at INR 399.

“Reliance is not attempting to replicate Stargate’s 10 GW ambition but to counter it with a different playbook. Its strategy is to localise compute at a gigawatt scale inside India, insulate it with self-supplied green energy, and distribute it through Jio’s digital grid,” Sanchit Vir, CEO of Greyhound Research, told AIM. 

This sovereign-first model directly addresses India’s compliance needs under the DPDP Act and resolves the chronic scarcity of affordable GPU capacity. 

Similarly, Rajgopal said that Reliance Intelligence will not have a role to play globally. “It will be focused on delivering value to users within India. Meta & Google have their own established agenda globally and do not need Reliance to play a role.”

Vir said Stargate is built to safeguard OpenAI’s product advantage. “It underwrites successive GPT generations and powers premium API, subscription, and enterprise sales,” he said, adding that the infrastructure itself is more of an enabler than a retail product. 

In contrast, Reliance Intelligence seeks to monetise across layers — from sovereign cloud and GPU-as-a-service through its Google Cloud partnership to packaged AI solutions with Meta’s Llama, and distribution via Jio’s connectivity, retail, and consumer platforms.

Notably, Reliance launched io AI Cloud 2.0, which offers AI-powered features beyond storage, allowing users to search photos, videos, and documents in their own languages using voice commands. The company also introduced an AI voice assistant named Riya, integrated with JioHotstar.

Rajgopal said the success of ChatGPT has given OpenAI access to large amounts of capital, enabling it to build the infrastructure needed to process vast datasets. “Through the Stargate project, OpenAI is deploying infrastructure in Abilene in the US, Abu Dhabi in the UAE, and has also announced sites in Norway and India,” he explained, adding that Stargate is a long-term investment aimed at powering the next generation of AI.

What About Indian LLMs?

On the LLM front, instead of building LLMs from scratch, Reliance announced a joint venture with Meta to build and scale enterprise AI solutions for businesses in India and select international markets. Under the agreement, the two companies have committed an initial investment of approximately INR 855 crore (US$100 million) to capitalise the venture. RIL will contribute 70% of the capital, while Meta will provide the remaining 30%.

Reliance’s use of Meta’s expertise remains an open question. Since Llama 2, Meta has seen limited success with LLMs and is now concentrating resources on its Superintelligence Labs to develop a frontier model.

Rajgopal said Reliance should have invested in developing the core technologies and capabilities needed for Sovereign AI in India. Instead, he noted, the company chose to focus on commercial ventures—likely to be successful given Jio’s access to millions of Indian users.

He added that sovereign AI is being built by small Indian companies with very limited resources, unlike projects such as Stargate. “Our ‘India moment’ will come from an innovation like DeepSeek. Constraints will bring out the best in people.”

The post Reliance Intelligence Ditches NVIDIA GPUs for Google TPUs appeared first on Analytics India Magazine.

Generated by RSStT. The copyright belongs to the original author.

Source