New Delhi: Private investments in India’s artificial intelligence (AI) infrastructure could double from last year’s $70 billion by the end of the ongoing financial year (FY26), Union electronics and information technology minister Ashwini Vaishnaw said on Friday.
Artificial intelligence infrastructure includes data centres and cloud computing servers with graphic processing units (GPUs).
“As far as AI infrastructure is concerned, so far we have already seen $70 billion in investments that are already on ground and in the execution phase. The way we see enthusiasm, we shouldn't be surprised if this number doubles by the end of the Summit,” Vaishnaw said.
Vaishnaw was addressing reporters ahead of India’s AI Impact Summit, slated to be held in New Delhi from 15 February to 20 February. In attendance at the Summit will be 14 heads of states, as well as at least 30 top executives including Stanford University’s Fei-Fei Li, Google DeepMind chief Demis Hassabis, former Meta chief scientist Yann LeCun, Nvidia chief Jensen Huang and others, the minister said.
This will be the fourth global AI Summit, after previous editions hosted by the UK, Korea and France.
India’s area of focus is to reach consensus on 10 actionable points on the development of AI around the world. Also on display will be a “bouquet of sovereign AI models,” which will be launched and demonstrated at the Summit next month.
Investment in AI infra
Shortly afterwards, on 14 October, Google announced an investment of $15 billion to build a 1GW data centre, in partnership with domestic conglomerate Adani Group. On 9 December, Microsoft chief Satya Nadella announced a $17.5 billion outlay, with primary focus on AI data centres. On the same day, Amazon announced it will invest $35 billion in India over five years, without specifying if a specific quantum was earmarked for data centres.
On 26 November, Reliance Industries’s data centre joint venture announced an $11-billion, five-year investment for a 1GW data centre. On 20 January, homegrown green energy firm Greenko Group’s subsidiary, AM Green, said it will invest $25 billion for a data centre of the same capacity, over five years.
Vaishnaw, however, added that as far as AI infrastructure is concerned, GPU-based capital-intensive facilities may not be the only way forward.
“AI around the world is progressing at a rapid pace, and today, various leading companies and their competitors are all working on newer forms of more accessible compute, as well as building small language models (SLMs) that can even run on a laptop. As a result, reliance on high-capex compute may not remain the same as it is today,” the minister said.
Noticed by world
“The way forward is sector-specific small AI models, and most industry experts I’m speaking with say that such small models are sufficient to provide for the needs of enterprise AI use cases.”
However, industry experts said access to AI infrastructure is likely to draw further investment into India through the year, though the deployment of that capital may become more measured over time.
“Capital will continue to come in, that much is clear. India is now firmly on the global data centre map, and the drivers that pulled money in during 2025 have not weakened. Enterprise demand continues to rise, hyperscalers are still expanding, and AI has pushed compute decisions into boardrooms rather than IT backrooms. At the same time, sovereign cloud has moved from policy language into procurement reality. All of this sustains investor interest,” said Sanchir Vir Gogia, chief executive of technology consultancy firm, Greyhound Research.
However, Gogia warned that investments and on-ground capital deployment may see a more nuanced approach.
“Where the confusion creeps in is the assumption that because 2025 saw an unusually large wave of announced investments, 2026 must automatically match or beat it on the same headline metric—that is not how infrastructure cycles normally behave. A large portion of last year’s capital surge was driven by signalling, big-campus declarations and long-horizon projects. Such announcement-heavy years are difficult to repeat back to back,” he added.
