‘AI can’t be just a bubble. Deep investments going into building infrastructure’: Google exec as company commits $15 billion to India data centre hub

The executive’s comments followed the Mountain View, California-based company’s announcement of a $15 billion investment to build data centre capacity for a new AI hub in Andhra Pradesh. The high-profile event in Delhi was attended by Finance Minister Nirmala Sitharaman, Minister for Railways, Information & Broadcasting, and Electronics & Information Technology Ashwini Vaishnaw, Andhra Pradesh Chief Minister N Chandrababu Naidu, and Google Cloud CEO Thomas Kurian.

The investment will be rolled out over the next five years, and the Visakhapatnam data centre will become Google’s largest AI hub in the world outside the United States.

‘Very real-life impact’

Tech heavyweights like Google are investing heavily in artificial intelligence infrastructure and operating AI models, even though it may take years to recoup these investments.

Alphabet’s Google, one of the top cloud service providers alongside Microsoft and Amazon’s AWS, is building large AI data centres and expanding infrastructure deployments, continuously adding more AI compute capacity to meet growing demand and user adoption from both consumers and enterprises. At the same time, major tech companies are caught in a race to maintain dominance and are increasingly seeking new ways to generate revenue.

Karan Bajwa, President of Google Cloud, Asia Pacific. (Image: The Indian Express/ Anuj Bhatia)

The cost of data centre spending — the giant buildings full of graphical processing units (GPUs), racks, and servers used by large AI firms to generate responses and train models — probably accounts for a major portion of expenses.

In fact, according to estimates, a little more than half of the cost of a data centre is attributed to GPU chips, which make up over 60 per cent of the total cost. The rest consists of cooling and energy costs, with only a relatively small component devoted to the actual construction of the data centre building.

Wall Street and investors, however, are increasingly concerned about how they will get their investment back — and when. Silicon Valley, on the other hand, continues to argue that AI could bring transformative changes, unlock opportunities that aren’t yet visible, and boost productivity across the economy.

“I would say there is a very real-life impact that has been created with AI,” Bajwa said.

Another senior Google executive added, “When we release a new model, for example, within a few hours, we are hitting upwards of a trillion tokens being processed in and out. That just shows how many people are consistently eager to use these new models as soon as they become available.”

Other issues

One of the biggest lingering concerns is the stress these data centres place on local power grids, and the significant water consumption they require. Data centres that power artificial intelligence consume immense amounts of water to cool overheated servers, and indirectly through electricity generation.

With local communities in the US already pushing back against the construction of new data centres, experts say companies may increasingly offshore data centre operations to countries like India.

US data centre demand could triple by 2030, according to McKinsey. By 2030, data centres could account for more than 14 per cent of the United States’ total power demand.