India, a late entrant to the AI party, is making efforts to make up for the lost ground to be in the top league. Abhishek Singh, Additional Secretary in the Union Ministry of Electronics and Information Technology (MeitY), said that the country would have at least 50,000 graphic processing units (GPUs) in the next 2-3 years, giving AI start-ups the much-needed firepower. GPUs are at the heart of the computation backbone for AI applications.

“We are lagging in computing power and research papers. The government itself commits for 10,000 GPUs, while Yotta announced plans to build a 25,000-GPU clusters and Tatas and Jios had announced plans to build 50,000 GPUs. We don’t know their timelines but we should be reaching a minimum of 50,000 GPUs in the next 2-3 years,” he said.

“The response to the RFPs for establishing computing infrastructure has been very good. As many as 50 companies showed interest. That should address the computing infrastructure issue,” he added.

Talking to reporters on the sidelines of the two-day Global AI Summit that began here on Thursday, he said that the Mission launched an application development initiative where 14 problem statements were identified across the five themes of education, agri, healthcare, flood forecast, climate change, and governance.

“For each problem statement, we announced a prize money of₹1 crore. The AI Mission has been approved. It backs all the seven key pillars including computing, foundational models, datasets, and start-ups. As many as eight out of 200 start-ups that submitted ideas to address some key challenges have been identified for funding support,” he said.

Highlighting the seven pillars of the country’s AI Vision, he said a key focus of the mission is to improve the country’s computer infrastructure and provide affordable computing resources to start-ups, researchers, and students.

Though the country has about 5 lakh datasets, not many of them are usable (by AI companies), he said. “Many of these datasets are not AI-usable due to challenges related to data standardisation, privacy, and accessibility. To address this, a data management framework is being worked out. We will be working with various departments to make relevant data usable,” he said.