CHENNAI: As artificial intelligence reshapes economies and governance worldwide, India continues to lag behind major powers in one critical area: access to high-end computing infrastructure. Despite its reputation as a technology and manufacturing hub, Tamil Nadu is no exception.
A white paper released this month by the Office of the Principal Scientific Adviser to the Government of India has flagged a widening gap in the country's AI readiness. While India generates and hosts close to one-fifth of the world's data, it accounts for only about 3 per cent of global data centre capacity and less than 5 per cent of AI-optimised computing power.
"The shortfall is increasingly becoming a bottleneck for domestic AI research and deployment. The contrast with global leaders is stark. The United States and China together command more than 70 per cent of the world's advanced AI compute resources, enabling them to train large language models domestically, reduce reliance on foreign platforms and retain strategic control over data-intensive technologies. India, by comparison, remains heavily dependent on overseas cloud service providers for compute-heavy AI workloads, a dependence that raises concerns over cost, data sovereignty and long-term competitiveness," the paper noted.
According to the white paper, within the country, AI infrastructure is unevenly distributed. Mumbai and Navi Mumbai together account for over a quarter of India's operational data centre capacity, driven by early policy incentives and strong subsea cable connectivity. Despite Tamil Nadu's long-standing strength in engineering talent, higher education and electronics manufacturing, Chennai contributes only about 13 per cent of national capacity, according to industry estimates.
"Tamil Nadu has the ecosystem, universities, skilled engineers and industrial demand, but the absence of large-scale, affordable computing is holding it back," a senior researcher at IIT Madras told DT Next, requesting anonymity as discussions with government agencies are ongoing.
"Many research ideas never progress beyond the prototype stage because access to GPUs is unpredictable and expensive," the researcher said.
The impact is already visible across the state's innovation landscape. Startups and academic institutions often struggle to secure sustained access to high-performance graphics processing units, forcing them to either cap the scale of their models or move workloads to foreign cloud platforms. This has limited progress in areas where locally trained AI systems could have a direct social and economic impact, including medical imaging, precision manufacturing, crop analytics and Tamil-language digital applications.
"AI models built on Indian health data or regional languages cannot indefinitely depend on infrastructure located abroad. When compute access is restricted, innovation becomes centralised, and solutions lose local relevance," said G Karthikeyan, a Bengaluru-based AI consultant who works with early-stage startups.
Both the Centre and the State have begun to acknowledge the challenge. The Union government's IndiaAI Mission is expanding a national GPU pool and offering subsidised compute access to startups and researchers through a centralised portal. Tamil Nadu, for its part, has rolled out a data centre policy that ties incentives to renewable energy usage, banking on its leadership in wind and solar power to attract energy-intensive facilities.
However, researchers and industry leaders argue that policy intent must translate into scale and accessibility. "Building more data centres is only part of the solution. The real issue is democratising access. Unless compute and datasets are treated as shared public infrastructure, AI capability will remain confined to a few metros and large firms," the IIT Madras researcher said.
For Tamil Nadu, experts say the next phase must involve pushing AI infrastructure beyond Chennai. Cities such as Coimbatore, Madurai and Tiruchy - home to universities, MSMEs and emerging startup clusters - will need to be integrated into national AI platforms if the state is to convert its talent advantage into real technological leadership.