India AI data centres have attracted more than $300 billion in investment pledges within the past year, according to a new report from the Rhodium Group. The report, written by Ankit Tiwari and Reva Goujon and titled “Hyperscaler Heaven? A reality check on India’s AI ambition and geopolitical alignment”, says operational capacity could triple by 2030, and could grow sixfold. The demand side of India AI data centres looks far weaker than the supply side, and that gap is the report’s main finding.
The Hindu BusinessLine reported the findings on 2 November 2026. The report works through the data centre buildout, the sovereign model strategy, the IT sector’s AI services work and India’s position in the technology competition between the United States and China. Its conclusion is plain: India AI data centres are being built fast, and much of the demand for them still comes from outside India.
India AI data centres and the demand gap
Graphics processing units supplied at subsidised rates under the IndiaAI Mission are running at roughly 20 per cent use, the report says, even though they are priced well below the market rate. Yotta, India’s largest GPU-as-a-service operator, gets an estimated 80 per cent of its revenue from foreign clients. Yotta has also agreed to distribute capacity through Nvidia’s global compute marketplace, DGX Cloud Lepton. Domestic enterprise AI workloads continue to run through global cloud providers rather than India AI data centres.
State governments backed India AI data centres with land, power waivers and capital subsidies, and they justified those incentives on employment grounds. The report records a wide gap between those promises and what data centres deliver in practice. Andhra Pradesh’s IT Minister cited 2 lakh jobs from one project, while the state’s own agency put direct operational positions at 200. Construction work peaks during the build and then disperses. Maharashtra revised its green energy incentive framework within five months of putting it in place, and Tamil Nadu and Karnataka have announced similar revisions.
India AI data centres and the alignment question
India has said it is not choosing sides in the technology competition between the United States and China. The Rhodium report argues that the choice has already been made at the level of infrastructure and regulation, whatever New Delhi says in diplomatic terms. India has aligned its semiconductor export control list with United States controls. It has adopted trusted-source requirements for telecom networks that exclude Chinese equipment. It has also taken positions in international spectrum allocation forums that broadly track the United States-led coalition rather than China.
China has responded, the report says. Beijing has recalled Chinese engineers from Foxconn plants in India, restricted tunnel boring machine exports and used rare earth controls in a selective way. The report reads those steps as signals that Beijing will not passively accept India’s role as an alternative to Chinese manufacturing.
India AI data centres and the services opportunity
AI labs backed by the IndiaAI Mission have put their effort into large language models, with Sarvam AI’s 105B parameter model the most prominent result so far. The report says the Indic-language case for such models is sound, since models trained natively on Hindi, Tamil or Bengali process queries more efficiently than models that translate on the fly. Its doubt is about market size rather than technical merit. The domestic market for Indic-specific AI products is not large enough to fund frontier model development on its own, and better capitalised foreign labs, American and Chinese, are closing the Indic performance gap.
The report is more positive about services work around India AI data centres. TCS and Infosys reported annualised AI services revenue of 7.5 per cent and 8 per cent of total revenue respectively in their most recent results, growing at double-digit rates, while overall IT revenue growth hit a six-year low. The mechanism the report points to is the forward-deployed engineer, a software professional who works inside a client organisation to fit AI deployment to its systems, workflows and data. An entry-level forward-deployed engineer based in India costs under a tenth of the equivalent role in the United States.
A 2025 study from MIT found that 95 per cent of generative AI pilots at large organisations fail because of lags in integration and knowledge gaps, rather than model underperformance. The report argues that Indian IT firms have built up that client-specific knowledge across decades of enterprise work, covering legacy systems, data architecture and workflows. Infosys puts the total addressable AI services market at $300 billion to $400 billion by 2030. The report’s point is that India AI data centres may matter less for jobs and domestic demand than the services work that helps clients put AI to use, and that policy for that services layer is thinner than policy for the buildout itself.