India faces AI development hurdles as global compute access tightens amidst rising concentration

As AI capabilities become reliant on large-scale processing power and data access, India’s ability to foster a broad-based ecosystem faces significant challenges from market concentration and upfront costs, prompting calls for policy reforms to democratise innovation infrastructure.

The older idea of a digital divide focused on internet access and device ownership. The new divide in artificial intelligence is broader and more strategic: who can afford the compute, model access and data needed to build and deploy advanced systems. That shift matters because AI capability now depends less on basic connectivity alone and more on whether firms, universities and start-ups can reach large-scale processing power, high-quality datasets and frontier models. The OECD has warned that these inputs are increasingly central to AI market power, while the World Bank has said low- and middle-income countries face serious constraints in obtaining the compute required to train and run models at scale.

This has clear implications for technological concentration. When a small number of companies control cloud infrastructure, proprietary data and foundational models, they can reinforce first-mover advantages and lock in users and developers. The OECD says such concentration can weaken competition in downstream markets, making it harder for new entrants to build viable products. The AI Now Institute has made a similar point, arguing that compute is now a critical bottleneck in the AI supply chain. In practice, that can turn AI from a general-purpose technology into an arena shaped by a few large gatekeepers.

For India, the risk is not only market concentration but also dependence. Fortune India has reported expert concerns that rising upfront costs, scarce talent and access gaps could limit the country’s ability to develop a broad-based AI ecosystem. A study by the Competition Commission of India has flagged steep barriers for start-ups, including weak access to data, expensive compute and financing gaps. If those obstacles persist, India may import more AI than it creates, weakening digital sovereignty and narrowing the space for domestic innovation.

The innovation cost is significant. Start-ups and smaller research teams often drive adaptation for local languages, sector-specific use cases and public-service applications. If they cannot obtain affordable compute or trusted datasets, they are forced to build on top of dominant platforms rather than compete with them. That can reduce experimentation, slow product diversity and make AI systems less responsive to Indian conditions. In a country as diverse as India, this would be a serious loss, because many high-value use cases depend on local language data, informal-economy patterns and context-specific deployment.

Building an inclusive AI ecosystem in India will require policy designed around access, competition and capability. First, India should expand shared compute infrastructure through public or public-private facilities that start-ups, universities and smaller firms can use at transparent prices. Second, open and interoperable data frameworks should be strengthened, especially for non-sensitive government and sectoral datasets. Third, support for indigenous and open models can reduce dependence on closed foreign systems and lower entry costs. Fourth, investment in AI skills, research funding and testing sandboxes is needed so that access to tools is matched by human capability. Finally, competition policy must prevent control of the AI stack from hardening into durable market power. The goal should be to treat compute, models and data as broad innovation infrastructure, not as privileges available only to a few firms.