In the next application phase, Google is thinking about contributing its Tensor Processing Units (TPUs) to the Rs 10,000 crore IndiaAI project.
Oracle is investigating involvement as well by providing assistance with computing infrastructure.

Google To Supply Tensor Chips in The Next Phase of IndiaAI Mission
The Ministry of Electronics and IT (MeitY) is in talks with both businesses about being appointed to the IndiaAI mission.
Due to delays in internal and legal approvals, Google was unable to take part in the initial round of proposals for computing infrastructure.
For large-scale AI model training and inference, Google’s TPUs—AI accelerators made for cloud AI workloads and data centers—compete with Nvidia and AMD’s GPUs.
Although Google designs its TPUs, third-party semiconductor foundries, mainly Taiwan Semiconductor Manufacturing Company (TSMC), produce them.
The government will pay 40% of the cost of a shared computing facility that will be utilized by researchers, startups, and academic institutions as part of the IndiaAI mission.
Rs. 4563 Crore Set Aside For AI Computing Infrastructure
Under the mission, Rs 4,563 crore has been set aside expressly for AI computing infrastructure.
Ten bidders offered to supply 18,693 GPUs for the IndiaAI mission in the first round of bidding.
E2E, Jio Platforms, NxtGen Datacenter, CtrlS, and Locuz Enterprise Solutions were the lowest-priced bidders.
Yotta, Tata Communications, CMS Computers India, Orient Technologies, and Vensysco are among the other bidders who must match the lowest bidder’s price in order to be empanelled.
The largest capacity, 9,216 GPUs, was proposed by Yotta.
To help startups access the available GPUs, the government is creating a compute access portal.
High expected demand is indicated by the expectation that bids for an additional 20,000 compute units will be invited in the next round of proposals.
Twelve,896 Nvidia H100, 1,480 H200, 512 AMD Mi325X, and 230 AMD Mi300X GPUs were among the 18,693 GPUs available in the first round, with additional models coming from Intel and AWS Trainium.
Higher precision AI compute units cost Rs 150 per hour, while the average compute unit rate under the IndiaAI mission is Rs 115.85 per hour.
In comparison to market rates, the government has obtained an average 42% discount on AI computing services.
A 47% discount is being offered on higher precision AI compute units, which are essential for training foundational models.
The empanelment process will continue to be open, according to government officials, so that additional entities can join at a later time.
To acquire more computer units for the IndiaAI mission, a new notice will be released shortly.