The man who was leading Infosys when it had backed OpenAI all the way back in 2015 believes that India still has a shot at joining the AI race.
Vishal Sikka, former CEO of Infosys and an advisor to OpenAI during its founding, has delivered a powerful message about India’s potential in artificial intelligence: the race is far from over. Speaking about the nation’s AI strategy, Sikka challenged the prevailing notion that India should abandon efforts to build foundational models and instead focus solely on applications. His argument is particularly striking given his unique vantage point—he led India’s largest IT services company during its early investment in what would become the creator of ChatGPT.
“India is large enough, broad enough, and deep enough and important enough that we should do all of that,” Sikka said, emphasizing the country’s capacity to pursue both foundation models and applications. “It is incredibly important for the future of the world that India gets this right. And the good news is that there is still a long road ahead.”
Sikka’s optimism stems from what he sees as fundamental inefficiencies in current AI technology. “We haven’t done this for the last three years, but the AI technology today is incredibly flawed. It is incredibly expensive. It is incredibly rare in the hands of a chosen few people. There’s a long journey ahead,” he explained.
To illustrate the technological gap that remains, Sikka offered a striking comparison between human and artificial intelligence efficiency. “I’m sitting in this room with six light bulbs. Each one of these is a hundred-watt light bulb. That is how much energy we consume in a day. Our entire body, 2000 calories that we eat, is basically one of these light bulbs and the brain, our entire intelligence, our nervous system, takes about 20 watts out of that,” he said.
The efficiency gap is staggering. “According to the experts, it is somewhere between 12 to 18 orders of magnitude more efficient. So 10 to the power of 12 to 10, to the power of 18, at least a trillion times more efficient. There are 12 zeros that need to be removed,” Sikka noted. Even accounting for rapid advances through Moore’s Law, he argued, “there is a huge space ahead.”
Beyond efficiency, Sikka pointed to fundamental gaps in current AI systems. “We don’t have understanding of the physical world. We don’t have an understanding of activities or grounding of concepts and all of these fundamentals that are still missing,” he said.
This technological reality led him to his central argument: “So it is not like we lost. It is a very, very bad idea, a very wrong idea to abandon that and let other people build. We have to build our own foundation models.”
Sikka dismissed the notion that building foundation models is beyond India’s reach. “Andrew Ng at Stanford actually teaches a class on how anybody can build a foundation model for God’s sake. Why should we give that up? We’re one and a half billion people. It’s not like some preordained people have been asked by God to build foundation models and we are not one of them.”
He advocated for a comprehensive approach: “We have to build our own foundation models and we have to build the best, greatest, widest collection of applications on top of them—both of them.”
Sikka also highlighted India’s unique advantages in the AI race. “All the languages, the work that we have in our manuscripts and our documents, the data that we have—these are all things that nobody else has. And now recently with the India Stack, we have incredible amounts of data that nobody else has access to. There is a lot that we can do that is different from everybody else.”
Sikka’s comments come amid an ongoing debate in India’s tech ecosystem about AI strategy. In 2015, when Sikka was CEO, Infosys was among the founding donors to OpenAI, contributing to a collective $1 billion pledge alongside Elon Musk, Sam Altman, and others. At the time, Sikka wrote that their wish was for “the OpenAI team to do unfettered research in the most important, most relevant dimensions of AI, not limited to just identifying dancing cats in videos, but to creating ideas and inventions that amplify our humanity”.
However, Sikka’s vision stands in contrast to views expressed by another Infosys co-founder Nandan Nilekani. Nilekani says that these models will ultimately become commoditized, and it might not be worth spending billions of dollars to build them. He believes it might be better to build AI applications on top of these models, because that’s where the real value might lie. But given how early we might be into the AI revolution and how future technologies might end up being built on these foundational models, like Sikka says, India would do well to develop this expertise in-house.