AI IS ERUPTING IN INDIA / THE ECONOMIST

A fresh perch

AI is erupting in India

American firms are piling on users—and sucking up mountains of data

Illustration: Juanjo Gasull

Sam Altman is bullish about India. 

The co-founder of OpenAI, the firm behind ChatGPT, says the country’s adoption of artificial intelligence (AI) has been “unmatched anywhere in the world”. 

India is already OpenAI’s second-largest market by number of users and could soon be its biggest. 

In August OpenAI launched a cheaper version of its chatbot tailored for Indian users. 

It plans to open an office in New Delhi later this year. 

Mr Altman is himself due to visit India at the end of this month. 

According to Bloomberg, a news organisation, he may use the trip to unveil plans to open a mammoth data centre there.

Other tech firms are just as keen. 

In January Microsoft, which is also an investor in OpenAI, pledged to spend $3bn over the next few years expanding its AI infrastructure in India. 

In August Google and Meta both announced partnerships with Reliance Industries, the conglomerate run by Mukesh Ambani, India’s richest man, to build data centres and push the use of AI in Indian businesses. 

In July Perplexity, a startup hoping to dent Google’s dominance in search, made the boldest move: it struck a deal with Bharti Airtel, one of India’s biggest telecoms firms, to provide its AI service (usually $240 a year) free for a year to all 360m Airtel customers. 

According to Sensor Tower, a market-intelligence firm, downloads of Perplexity in India soared by almost 800% month-on-month after the Airtel tie-up, compared with gains of 39% and 6% for ChatGPT and Gemini (see chart).

For Indians, the battle promises cutting-edge AI tools at extremely low cost. 

For AI firms, the payoff may lie less in revenue than in reach: the chance to lock in hundreds of millions of users, and the torrents of data they create.

The opportunities in India are staggering. 

The country has around 900m internet users; only China has more. 

But unlike China, India is open to American tech firms. 

Google’s Android powers over 90% of the country’s smartphones. 

WhatsApp, owned by Meta, has more than 500m active users. 

India’s e-commerce sector is dominated by American giants: Flipkart, owned by Walmart, and Amazon. American firms that already have millions of users for products such as search, shopping or messaging have a big head start when it comes to driving uptake of AI services. 

And Venugopal Garre of Bernstein, a research firm, says tech firms that do well in India can add users at a speed and scale that few other markets can match.

Yet wringing profits from this vast user base will be an altogether trickier task. 

Most tech firms charge less in India than they do elsewhere. 

For example, Netflix, a video-streaming service, costs as little as $1.69 a month in India, compared with $7.99 in America. 

For cloud services with a low marginal cost, this is no great sacrifice. 

But running AI queries is expensive. 

Processing costs for typical users currently hover at around $0.07 per million “tokens” (the units of data processed by AI models) and the response to a single query can run to hundreds or thousands of tokens. 

That expense is the same whether the user is in Bangalore or the Bay Area.

Dmitry Shevelenko, Perplexity’s chief business officer, admits that giving away its service for a year is costly, but says the company has time to “prove ourselves” and convert users into paying customers. 

India, he notes, shows some of the highest engagement in the countries where Perplexity has run trials. 

Mr Shevelenko predicts that within five years India will be a “compelling subscriptions market”. 

That remains to be seen. 

Though India is vast in users it is small in paying customers, as many other firms have discovered.

Fortunately, subscribers are not the only draw. 

India’s internet user base is huge, diverse and messy. 

It encompasses people from many different language groups; it jostles together people who are very wealthy with some who are extremely poor. 

An executive at a leading AI model-maker describes the country as an ideal “testing ground” for new products, and for observing usage at scale. 

Already firms are noting that Indian users prefer to speak to their AI models, rather than to type out their queries—perhaps because many Indian internet users cannot read or write.

AI firms have mostly already scoured all large public data sets for information they can use to improve their models. 

So Indian users are a valuable new source of material to crunch. 

The country has built a robust digital backbone known as the “India Stack”, which bundles together state-led services such as a biometric ID system and a digital payments system. 

That infrastructure has brought online hundreds of millions of people who were previously locked out of many internet services. 

Their queries provide precious and novel real-world data that can help fine-tune models.

India’s regulatory regime helps. 

Konark Bhandari of the Carnegie Endowment for International Peace, a think-tank, notes that rules do not restrict companies from transferring data across borders. 

There is, he adds, “nothing on the books” to prevent data collected in India being used to train models overseas. 

As other sources of fresh data run dry, the potential fire hose in India is a prize few AI firms can ignore.

Indian consumers mostly welcome the arrival of foreign AI firms. 

But some worry about the impact on the country’s big IT-services firms. 

And many fear long-term dependence on American platforms. 

That anxiety has only grown since President Donald Trump imposed steep tariffs on India. 

Mr Garre argues that American firms, with more money and stronger infrastructure, could “kill India’s prospects” in AI by deterring investment in local startups. 

That, he warns, could echo previous periods of fast technological innovation—when Indian companies mostly settled for rigging up peripheral services while their American peers constructed the core platforms. 

Though the country has one of the largest populations of developers, the number of AI researchers is low.

Mr Altman offers a cheerier vision. India, he has said, can be “one of the leaders of the AI revolution”. 

The question is what sort of leader it will choose to be: one that dominates by dint of its massive user base, or through creating its own technology?