It's a muggy March afternoon in the suburbs of Mumbai and a group of men anxiously hover around the back door of a startup called Yotta Data Services. They walk, pause and worry. It's approaching midnight, 10 hours late, when a truck arrives with the precious cargo they've been waiting for: semiconductors from Nvidia Corp.
The company's products are highly coveted because they are essential to the development of artificial intelligence, the technology that has sent industries around the world into a frenzy. While companies like OpenAI and Google have invested billions of dollars in such chips in the United States, Yotta is making India's biggest bet yet on the promise of AI.
CEO and co-founder Sunil Gupta has leapfrogged the country's best-known conglomerates and tech players in part because of the relationship he has forged with Jensen Huang, the famed CEO of Nvidia. Yotta is expected to participate in Nvidia's developer conference on Monday in California, an early example of AI's potential in markets beyond the United States.
“I'm ambitious, I'm hungry,” said Gupta, 52. “I'm willing to bet on the future of AI.” Yotta's strategy is to offer high-performance computing capabilities from data centers in India so that corporations, startups and researchers in the country can develop their own AI services.
Nvidia's chips, the most advanced on the market, are essential for training large language models and building applications like OpenAI's ChatGPT and Microsoft Corp.'s coding assistant, GitHub Copilot. Gupta believes it has an advantage over cloud computing services outside the country due to latency issues and promises to offer the cheapest access to Nvidia AI chips in the world.
It is even considering allowing Indian startups with tight budgets to give it equity instead of cash. The demand is on your side. The global AI market is projected to grow from US$168.5 billion in 2022 to more than US$2 trillion in 2032, according to a report by Spherical Insights & Consulting.
“This is a gold rush,” said Stacy Rasgon, an analyst at Sanford C. Bernstein. “We're still in the early days of AI and companies simply can't buy enough of this stuff.”
The new era got off to a rocky start this month in India. The country's customs officials were baffled by the unusually high value of the Nvidia chips that Yotta had purchased, prompting requests for additional paperwork and bureaucratic approval. Back at his data center on the outskirts of Mumbai, Gupta walked the marble floors of the lobby for most of the day, working the phones to unlock their chips.
The delivery truck finally stopped and workers unloaded the first of the more than 4,000 H100 chips that Yotta ordered from Nvidia. The beefy graphics processing units, or GPUs, cost between $30,000 and $40,000 each and are called Hoppers in a nod to computing pioneer Grace Hopper.
Childbirth was a religious experience for Gupta, literally. A priest adorned the boxes with vermillion red markings and strings of yellow chrysanthemum flowers, while hymns in ancient Sanskrit filled the night air. A drone with a camera recorded Gupta symbolically crushing a coconut on the ground near the truck. “It's a dream moment,” she said, amid bursts of party pops.
Yotta's haul of Nvidia chips, which will hit around 20,000 in June, isn't huge by global standards. Tech giants like Microsoft Corp. buy them by the tens of thousands, and Mark Zuckerberg of Meta Platforms Inc. said he aims to get 350,000 H100s by the end of the year. Still, Nvidia's supply falls far short of demand, so CEO Huang has to calibrate allocations as corporate titans and heads of state push for allocations.
India is receiving special attention. In September, Huang met Prime Minister Narendra Modi and said he would prioritize any requests from the country's data center operators. “You have the data, you have the talent,” Huang said at the time. “This will be one of the largest AI markets in the world.”
The next day, Gupta received a call from the Nvidia team asking if he could meet the CEO in the western city of Pune. Although it was already late and the meeting would be the next morning, Gupta quickly agreed. He got in his car and drove three and a half hours through the night to attend the talk. It was a demonstration that Yotta would go further.
Gupta has a lot of good faith in the field. He has worked for decades in data center businesses and co-founded Yotta in 2019 with the backing of real estate billionaire Niranjan Hiranandani. As a cloud computing operator, Yotta gives companies like Wells Fargo & Co. access to data storage and computing power that they can scale up or down as needed, without purchasing or installing their own hardware.
Tata Group and Reliance Industries Ltd., two of the country's largest conglomerates, also plan to develop artificial intelligence infrastructure, but have not yet ordered Nvidia's most advanced chips.
An Nvidia spokeswoman declined to comment on the details of Yotta's order, noting that more will be revealed this week. Gupta will speak at Nvidia's GPU Technology Conference and was told that Huang will talk about Yotta during his keynote speech on Monday.
One reason for this attention is the global imbalance in AI. If technology has the potential to transform virtually every industry, as Huang and Microsoft CEO Satya Nadella argue, then countries like India, Indonesia and Turkey are at risk of not having access. In India, that could hamper scientific research, startup development and, more broadly, Modi's ambitions to create a technological superpower. “GPU disparity” is an increasingly popular term for the dilemma.
“Countries that do not have their own AI infrastructure and models will unfortunately lose the AI race,” said Umakant Soni, co-founder of a nonprofit AI and robotics research park called ARTPARK. Gupta sees a clear need to develop AI models built in India, trained in local languages and cultural diversity. “India needs sovereign AI, India needs sovereign models,” he said.
Geopolitics is helping his case. Rising tensions between the United States and China have led the Biden administration to impose extensive controls on the export of technologies to its geopolitical rival, including Nvidia's own H100 chips that Yotta is purchasing. Cloud providers in the Middle East have also come under scrutiny after a key US lawmaker urged the Commerce Department to investigate the Chinese connections of Abu Dhabi-based artificial intelligence firm G42.
Gupta believes he can supply Indian customers and others in Asia and the Middle East. Yotta already has half a dozen data centers in four Indian cities and a new one opening in northeast India. The businessman named his company after the number eight in ancient Greek, which represents a septillion.
“India is trying to catch up,” said Nruthya Madappa, partner at venture capital firm 3one4 Capital. “But because of the talent base, we see that the recovery will be very, very quick.” The seven-story data center on the outskirts of Mumbai is surrounded by electric fences, equipped with 850 cameras and includes seven layers of security. The massive diesel storage tanks contain enough fuel to run the facility for 48 hours if the power goes out.
Gupta's partnership with Nvidia requires such rigid protocols, along with strict specifications to build the AI cloud business. He has sealed off the entire sixth floor of the facility for that purpose. An Nvidia team will arrive in the coming weeks to get the network up and running, with the goal of starting operations in mid-May. Gupta calls the first cloud service H100 Shakti, the Hindi word for power.
It says it has exhausted capacity for the day its network goes live and has a waiting list of companies from India and beyond. Gupta is already looking forward to the next delivery of Nvidia chips, more than 16,000 expected in June. However, he will do at least one thing differently: hire guards, since the value of the shipment could run into hundreds of millions of dollars.
“Security?! I never thought about that!” said Gupta. “A lot of people want this.”
