As global tech companies race to release ever larger artificial intelligence models, a more decisive contest is emerging behind the scenes. The countries that will lead the next decade of AI are not the ones with the most visible applications but the ones that control the engineering and infrastructure layers that make advanced systems possible, believes Omprakash Subbarao, Chief Executive of CORE Labs at the Indian Institute of Science. He argues that India’s long-term position depends on whether it can build the stack on which AI runs.
“The global AI competition is now less about who has the most impressive app and more about who has the research capacity to push the efficiency threshold in a new direction,” he says. India’s strength, he adds, lies in its deep mathematical and algorithmic talent, which he describes as the country’s “logic layer”. The challenge sits elsewhere. “We remain immature when it comes to engineering large-scale systems and optimising hardware for scaling purposes.”
That gap shows up most sharply in the lack of affordable, sustained access to high-end compute. For many Indian researchers, a single model-training experiment can take weeks. In leading global labs, similar jobs finish in hours. Subbarao calls this the largest tax on innovation. “Researchers stop iterating over ideas when it requires weeks to wait for a single training run. At other locations, this takes just a couple of hours.”
More and more institutions, including IISc, are now trying to close this gap by building capabilities that the private sector has been reluctant to carry. Their focus is on sovereign infrastructure, including specialised storage systems and AI accelerators that are designed for local needs. The aim is to reduce dependence on foreign platforms and create the conditions under which Indian researchers can experiment at scale.
“With an emphasis on the interaction between hardware-aware systems and data from their respective domains, we are able to transition from a tenant in the development of global technologies into an architect of our own infrastructure,” he says.
For young professionals and students, this trend also demands a change in how they think about their careers. Subbarao believes the next generation will need to engage far more deeply with the systems beneath the applications they use.
“They will have to gain a solid understanding of how data flows throughout a silicon-based system and through the software that delivers those same systems to an end user.” Instead of relying on high-level tools, he says, early-career engineers must learn to work on the “engine” that creates value.
Over the next decade, India’s ability to compete will depend on whether it can design and maintain the unique frameworks required for advanced AI while ensuring that its digital public infrastructure remains both autonomous and globally competitive. The race, in other words, may be won by those who build the foundations, not only those who run on them.