Budget 2025 fuels AI boom and nuclear push amid infra concerns
India's 2025 Budget has placed major bets on AI innovation and nuclear energy to meet soaring GPU-driven demands.
After last year’s Budget that overlooked artificial intelligence (AI), Finance Minister Nirmala Sitharaman unveiled an ambitious slate of initiatives geared towards the tech sector in Budget 2025.
Key among those is a Rs 10,000 crore Fund of Funds to support the startup ecosystem, and a deeptech Fund of Funds to spur next-generation of startups.
Notably, AI also received a nod.
According to estimates provided by the government, the India AI Mission's (IndiaAI) allocation has surged tenfold to Rs 2,000 crore from Rs 173 crore (revised estimates from FY24-25), with a substantial part of the allocation expected to go into funding the GPU compute project under the AI mission.
IT Minister Ashwini Vaishnaw recently outlined the nation’s bold LLM plan which leans heavily on computing infrastructure. He said IndiaAI also overachieved on procuring 19,000 GPUs than its original target of 10,000 through public-private deals, stockpiling a total of 18,600—including several NVIDIA H100 and H200 chips, along with AMD.
Push towards AI research
The Economic Survey 2024-25 flagged a few critical issues with AI, including its heavy resource demands, potential to displace workers, along with infrastructural challenges. It notes that India’s massive workforce and low per capita income could amplify job losses, stressing the urgent need for stronger institutions and enhanced skill-training programmes.
To address this, the FM announced that the government will set up an AI (Centre of Excellence) for the education sector, which will be established with an outlay of Rs 500 crore.
“I had announced three Centres of Excellence in Artificial Intelligence for agriculture, health, and sustainable cities in 2023. Now a Centre of Excellence in Artificial Intelligence for education will be set up with a total outlay of Rs 500 crore,” Sitharaman said.
Jameela Sahiba, Senior Programme Manager-AI at The Dialogue, believes that the budget has shifted focus from direct tech investments to digital facilitation, focusing on education, AI-driven advancements, and regulatory modernisation.
“The Union Budget 2025-26 signals a strategic shift towards digital facilitation rather than too many direct investments in technology initiatives. The budget also makes a strategic investment in a Centre of Excellence (CoE) on Education, reinforcing the role of technology in advancing learning and skill development,” Sahiba tells YourStory.
This comes days after the government, on Thursday, announced plans to develop its very own foundational model in less than a year to compete with likes of OpenAI’s ChatGPT, Google’s Gemini, and DeepSeek— the trending Chinese AI model that triggered a shakeup in the stock market, which led AI giants like NVIDIA to lose in billions.
“The recent tender for GPU hours, with a total compute budget of Rs 4563.36 crore under IndiaAI, has yielded impressive results. Around 10 players are offering 18,693 GPUs at an average bid price of Rs 115.85 ($1.34) per GPU hour, significantly lower than the global average of $2.5-$3. With a 40% subsidy from the government, the effective price will be Rs 69.51 per GPU hour,” Ankush Sabharwal, Founder and CEO, CoRover tells YourStory.
The minister compared China’s DeepSeek, which is reportedly trained on around 2,000 GPUs, to ChatGPT, which utilised 25,000 GPUs for its training.
“There is a lot of attention to research and development in the AI/ML areas in India’s premier technology institutions such as IITs and others. The government should allow R&D investments to be treated as assets and create the ability to depreciate the assets to allow for experimentation within the private enterprises,” says AS Rajgopal, CEO of NxtGen.
The GPU power problem
GPUs are the backbone of AI model training. Despite the GPU stockpile by IndiaAI, which will be provided to startups, researchers, and academia–experts have cautioned that India’s path forward hinges on aligning hardware capabilities with AI research needs, as data centre infrastructure necessary for large-scale AI training is still a roadblock.
Moreover, the soaring power and water requirements for AI-driven data centres could also strain the country’s resources, according to the Economic Survey 2024-25.
“The capability to support large-scale AI model training extends beyond mere GPU availability – it will have to consider needs such as data centre infrastructure, energy efficiency, cooling systems, and high-speed networking. The current GPU acquisition is a positive development, but continuous investment in these supporting infrastructures is essential to meet the escalating demands of AI R&D,” Karan Kirpalani, CPO, Neysa tells YourStory.
“No data centre in India is designed to handle 100KW of power per rack, it is commercially unviable if a traditional design is done for such high-power consumption. We are deploying H200SXM (NVIDIA) & MI325X (AMD) both are rated to consume about 10KW and 8 servers in each rack, amounting to rated power consumption of 80KW per rack. In reality, actual consumption is about 56% of the rated capacity,” Rajgopal said.
Kirpalani advises that Indian data centres should quickly invest in AI-optimised infrastructure with pre-configured GPU offerings for large AI workloads. He also highlights the importance of AI-ready data centers powered by sustainable, hybrid energy solutions to enhance efficiency in AI.
Betting on nuclear energy
The government has plans for AI infrastructure. Sitharaman announced plans to develop at least 100 GW of nuclear energy by 2047, a major step in India's energy transition efforts.
The minister stated a Nuclear Energy Mission focused on the research and development of Small Modular Reactors (SMRs) will be established, with an investment of Rs 20,000 crore, with the goal of having at least 5 indigenously developed SMRs operational by 2033.
The government’s renewed push for nuclear energy is partly geared toward meeting these power-hungry demands of GPUs.
“The government’s Nuclear Energy Mission and its Rs. 20,000 crore R&D push for Small Modular Reactors (SMRs) will accelerate the transition to clean energy solutions, helping the data centre industry achieve long-term carbon neutrality. This commitment to sustainability is essential for the future of data centres, enabling the development of a low-carbon infrastructure,” says Sridhar Pinnapureddy, Founder and CEO, CtrlS Datacenters.
Tech behemoths, including Amazon, Google, and Microsoft, are betting on SMRs to power data centres and AI, given their relative cost-effectiveness and quicker deployment compared to traditional nuclear power plants. These reactors are known to generate less power than existing ones, along with having various fuel types and cooling systems in their designs.
“The opening of the nuclear sector to private companies and the move towards smaller Modular Micro Reactors (MMRs) marks an important step in moving away from fossil fuels and diversifying energy sources. In the future, with the expected increase in data centres, nuclear energy will become essential, especially as smaller reactors are being developed worldwide,” says R. Ajai Chowdhry, Founder of HCL, and Chairman of EPIC Foundation.
Rajgopal argues that nuclear power plants offer a consistent and reliable output of energy, making them ideally suited for the demands of advanced computing.
“In the near future, we can anticipate the deployment of microreactors throughout the country, a crucial step in meeting the growing energy needs of our computing landscape. With these forward-thinking measures, AI service providers like us are well-positioned to drive digital transformation across industries and contribute to India’s vision of becoming a $5 trillion economy.”