India hosts 22 AI producers, but cloud services dominate value: BIS paper

India ranks seventh by aggregate AI-firm valuation, but has negligible value in AI models and only a thin presence in computing infrastructure, according to the BIS study

India is home to 22 firms producing artificial intelligence (AI) products and services, placing it alongside Canada and behind Japan (26), the UK (40) and Israel (42), according to a Bank for International Settlements (BIS) working paper, ‘The Geography of AI Firms’.

 

The study by Kumar Rishabh and Vatsala Shreeti, economists in the BIS's Monetary and Economic Department, maps 1,246 AI-producer firms across 32 economies, each valued at more than $500 million. The dataset was built using PitchBook data screened by a large language model and manually verified. Firms are classified into five layers: compute, cloud and related infrastructure, data tools, AI models and applications.

 

The US accounts for nearly 700 of these firms and China for around 250. India ranks seventh by aggregate AI-firm valuation, ahead of the UK and Canada.

 

The paper is significant at a time when the buzz around ‘sovereign AI’ is gaining momentum. Until now, there has been little systematic data on what “having an AI industry” actually looks like on the ground. This includes data such as how many genuine AI-producing companies a country hosts, which parts of the supply chain they occupy and how they compare globally. The BIS study fills that gap.

 

The composition is more revealing for India. Almost all of India's AI-firm valuation sits in the cloud and infrastructure layer. The authors note that this is largely an artefact of classifying information technology services majors such as Wipro and HCL Technologies in this category, since their businesses resemble systems integration and cloud consulting rather than the operation of cloud platforms. The authors concede that the interpretation is debatable. India has negligible valuation in AI models and only a thin presence in compute.

 

AI firms accounted for less than 6 per cent of total listed-company revenue in India in 2024, against roughly 13 per cent in the US and Korea. Their share of capital expenditure was below 5 per cent and had barely moved since 2010. India's AI-firm valuation is also skewed towards older companies: Firms founded after 2012 make up 27 per cent of the count but only about 2 per cent of the valuation.

 

In terms of investment behaviour, India is an outlier in two ways. It is one of only three economies — along with Canada and Sweden — that does not show significant home bias. Indian AI firms direct 41.7 per cent of their investment deals towards US-based targets, against 27.9 per cent domestically. India, along with Japan and the US, is also one of three economies where most deals do not remain within the investor's own supply-chain layer, suggesting that Indian AI firms buy capabilities rather than reinforce them. Applications accounted for 52.6 per cent of Indian deals and infrastructure for 15.5 per cent.

 

The paper's regressions find that venture-capital inflows, rather than bank credit, correlate with both the presence and number of AI firms in an economy, with bank lending being collateral-based and ill-suited to young, intangible-heavy firms. Diversification across layers tracks university-industry collaboration and venture-capital depth, and falls as manufacturing's share of gross domestic product rises.