The data centre sector is booming. It doesn’t need state subsidies.

Rather than subsidize an industry that is booming worldwide, policy could act as an enabler in areas that range from the low-cost supply of uninterrupted power to the training of people.

Data centres are hardware sites needed for shifting data and software services to the digital cloud in general and for the development and running of artificial intelligence (AI) in particular.

Developing AI is vital to India in a way it is not to many other countries, given the impact AI will have on information technology (IT) and IT-enabled services (ITeS), apart from all manner of human activity.

The IT and IT-eS sectors play a role that goes beyond their direct share in national output. As employers, they offer a path to social mobility for large swathes of first-generation college-goers and spur demand for a host of other industries.

They can count education, construction, transport, hospitality and entertainment among their indirect beneficiaries. They not only generate salaries, wealth and unicorns, they absorb much of the venture capital that flows into India and remain integral to the dynamics of future economic growth.

Globally, data centres are all the rage. Beyond serving near-term IT needs, they are expected to form a backbone for the evolution of today’s digital world into the AI age, in which India must actively participate or risk being left out as a mere user of foreign tools. But that does not call for a data-centre subsidy.

At the moment, demand is rising robustly for large installations that host processing power and data storage. How it drove up Oracle’s stock price—and Larry Ellison’s wealth—has been a clear sign of this boom.

In time, demand should attract enough data centres by way of supply without any state incentive needed to set them up. What we must avert, particularly, is an excessive build-up. Data centres guzzle power and our baseload may not expand fast enough to feed such ravenous IT infrastructure.

While renewable energy offers an answer, we must not find clean capacity failing to keep up with the sector’s appetite. Instead of back-end infra, we should focus resources on gaps at the innovative end of AI.

Success here requires us to raise the country’s R&D spend from its dismal level of 0.64% of GDP. An ambitious AI thrust would call for innovation in almost everything IT related, including high-end chip design and manufacture.

While public funds are going into various fields, chip-making included, fiscal resources are best used for basic research and development work—to plug a void left by private investors.

Morgan Stanley, as cited by The Economist, has forecast $2.9 trillion in AI investments to be made till 2028. Revenues of the top Western AI companies still add up to only $50 billion, though.

Even if sales rise at a dizzying pace, they may struggle to justify the money going into an IT infra frenzy. Many analysts reckon that AI hype has created an investment bubble that could burst, just as the dotcom one did in 2000.

The Indian market, while partly driven by data localization norms and other local requirements, would not be insulated if that happens. Private builders of data centres should be asked to venture forth without the support of public money.

This isn’t a sector that needs a special incentive to reach its ideal capacity.