In the Mumbai suburbs on a steamy March evening, a squad of men are standing guard at the rear gate of a startup company named Yotta Data Services. They pace, hesitate, and worry. Ten hours after the scheduled time, at around midnight, a truck bearing the much-needed cargo—Nvidia Corp. semiconductors—pulls up.
The reason the company’s products are so sought after is that they are crucial to the advancement of artificial intelligence, a technology that has sparked a global industry frenzy. Yotta is placing the biggest wager on the potential of AI in India to date, despite the fact that US businesses like OpenAI and Google have invested billions of dollars in similar chips.
The bond that Sunil Gupta, the chief executive officer and co-founder, has built with Jensen Huang, the celebrity CEO of Nvidia, has helped him get an advantage over the nation’s more well-known technological firms and conglomerates. Yotta is anticipated to be featured at Nvidia’s developer conference on Monday in California, serving as a prototype for AI’s possible applications outside of the US market.
Gupta, 52, declared, “I’m hungry and ambitious.” “I’m ready to wager on artificial intelligence’s future.”
Yotta’s plan is to provide high-performance computing power from Indian data centers so that businesses, startups, and researchers can create their own artificial intelligence services in the nation. The most sophisticated CPUs available are from Nvidia, which is necessary for developing software like Microsoft Corp.’s GitHub Copilot and OpenAI’s ChatGPT, as well as for training big language models. Due of latency concerns, Gupta believes he has an advantage over cloud computing services offered outside of the nation. He also promises to provide the most affordable access to Nvidia AI chips globally. He’s even thinking about accepting equity rather than cash from Indian firms that are operating on a shoestring.
Demand is in his corner. A report by Spherical Insights & Consulting projects that the global AI market would reach over $2 trillion by 2032, up from $168.5 billion in 2022.
Sanford C. Bernstein analyst Stacy Rasgon declared, “This is a gold rush.” “AI is still in its infancy, but businesses can’t get enough of this stuff.”
In India, the new age began this month with some difficulties. Yotta’s abnormally high purchase of Nvidia processors confounded the nation’s customs officials, resulting in requests for more documentation and regulatory permission. For nearly a whole day, Gupta walked around the lobby’s marble flooring in his data center located outside of Mumbai, contacting people to obtain the release of his chips.
When the delivery truck did arrive, personnel began to unpack the first of Yotta’s order of over 4,000 H100 chips from Nvidia. In honor of computer science pioneer Grace Hopper, the robust graphics processing units, or GPUs, are dubbed Hoppers and cost between $30,000 and $40,000 per.
For Gupta, the delivery was practically a holy experience. Ancient Sanskrit songs permeated the night air as a priest decorated the boxes with crimson vermilion marks and strings of yellow chrysanthemum flowers. Drone with camera attached captured Gupta breaking a coconut on the ground in front of the truck. He remarked, “It’s a dream moment,” as party poppers burst out.
By global standards, Yotta’s haul of Nvidia chips isn’t that great—it will reach approximately 20,000 by June. Tens of thousands being bought by tech behemoths like Microsoft Corp., while Mark Zuckerberg of Meta Platforms Inc. stated that he plans to acquire 350,000 H100s by the end of the year. Even still, CEO Huang must balance allotments as heads of state and giant corporations push for higher allocations due to Nvidia’s supply falling well short of demand.
India is the focus of particular attention. Huang told Prime Minister Narendra Modi during their September meeting that he would give national data center operators’ requests top priority. Huang stated at the time, “You have the talent, you have the data.” “One of the biggest AI markets in the world is going to be this one.”
The Nvidia team called Gupta the following day to ask whether he could meet with the CEO in Pune, a city in western India. Gupta agreed right away, even though it was late at night and the meeting would take place the following morning. He got into his car and drove to the conference for three and a half hours at night. It was an example of Yotta going above and beyond.
Gupta has a strong background in the industry. With the support of real estate mogul Niranjan Hiranandani, he co-founded Yotta in 2019 after working for decades on data center firms. As a cloud computing provider, Yotta allows businesses, such as Wells Fargo & Co., to obtain processing power and data storage that they can scale up or down based on demand without having to purchase and install their own hardware.
Although they haven’t yet placed an order for Nvidia’s most cutting-edge chips, two of the biggest corporations in the nation, Tata Group and Reliance Industries Ltd., also intend to develop AI infrastructure.
For access to unique material and analysis on the actions of billionaires, businesses, and markets, follow Bloomberg India on WhatsApp. Register by going here.
A representative for Nvidia declined to comment on Yotta’s order details, stating that further information will be made public this week. According to Gupta, who is giving a talk at the Nvidia GPU Technology Conference, Huang will talk about Yotta in his keynote address on Monday.
Unevenness in AI around the globe is one factor drawing attention. According to Huang and Microsoft CEO Satya Nadella, the technology might revolutionize almost every industry, so nations like Turkey, India, and Indonesia would be in danger if they don’t have access to it. That might hinder scientific research, the growth of startups, and, in general, Modi’s goals of making India a technological giant. The problem is increasingly being referred to as “GPU disparity.”
Umakant Soni, co-founder of the nonprofit AI and robotics research park ARTPARK, predicted that “countries who don’t have their own AI infrastructure and models will woefully lose the AI race.”
Gupta believes that creating AI models in India that are trained with regional languages and cultural diversity is imperative. He declared, “India needs sovereign models, and India needs sovereign AI.”
His position is aided by geopolitics. The Biden administration has imposed extensive limitations on the sale of technologies to its geopolitical rival, China, in response to growing tensions between the two countries. These controls include the H100 Nvidia chips Yotta is purchasing. Middle Eastern cloud providers have recently come under fire after a prominent US legislator pushed the Commerce Department to look into the Chinese ties of G42, an AI company based in Abu Dhabi.
Gupta believes he can supply consumers in Asia and the Middle East, as well as those in India. Yotta now operates six data centers in four Indian cities, and a new facility is set to launch in the northeast of the country. The business owner gave his company the ancient Greek number eight, which stands for one septillion.
Venture capitalist Nruthya Madappa, a partner at 3one4 Capital, remarked, “India is playing a bit of catchup.” “However, we see the catch-up being very, very fast because of the talent base.”
The seven-story data facility outside of Mumbai has seven layers of security, 850 cameras, and electric fences all around it. Massive diesel storage tanks provide enough fuel to keep the facilities operational in the event of a power outage for 48 hours.
Such strict rules and strict specifications are required for developing the AI cloud business as part of Gupta’s relationship with Nvidia. For that reason, he has completely closed off the sixth story of the building. A team from Nvidia will be arriving in the upcoming weeks to set up the network, with plans to begin operations by the middle of May. Gupta gives the inaugural H100 cloud service the Hindi word for power, Shakti.
He claims to have a waiting list of businesses from India and abroad and to have sold out capacity on the day his network goes live. With the next delivery of Nvidia chips—more than 16,000—scheduled for June, Gupta is eagerly anticipating it. Nevertheless, given the shipment’s value might approach hundreds of millions of dollars, he will at least hire guards.
