OpenAI CEO Sam Altman is in Delhi today for a public event, including a closed-door panel with IT Minister Ashwini Vaishnaw and other key AI stakeholders. His rushed visit to Asia comes amid OpenAI’s dispute with Indian media houses and the global rise of China's DeepSeek.

During his discussion with Vaishnaw, which was part of the closed-door panel, Fortune India has learned that Altman addressed the debate over whether large language models (LLMs) truly require high costs, especially considering DeepSeek’s success.

Altman highlighted how advancements in AI efficiency are driving both cost reductions and increased value creation. He emphasised that while the costs of AI models are decreasing, overall investment in AI will continue to rise as lower costs drive wider adoption and innovation.

“The cost of a given unit of intelligence seems to fall by about 10% each year. Moore’s Law predicted a doubling of transistors on a chip every 18 months, and that pattern persisted for decades. But the cost reductions in AI models are extraordinary. This means the world will need less AI hardware because costs are coming down, allowing for more applications. However, the total investment in AI will continue to increase,” Altman said.

“Models are still not cheap, but they are feasible. India should be a leader in this space, of course,” Altman added. He also expressed strong interest in collaborating with India on AI projects, noting that the country is now OpenAI’s second-largest market.

“Every year, there’s a tenfold reduction in cost. And innovation can come from anywhere—why shouldn’t it come from India? That’s the point. Our young entrepreneurs, startups, and researchers are focused on the next level of innovation, which will further reduce costs,” Vaishnaw said.

“Given the diversity we have here, if you can solve for India, you can solve for the world as well,” he added.

AI developers and startup founders are attending the discussion. The event was organised in complete secrecy with sources saying that the event venue was also shared only later in the night. There are other panellists as well attending the event.

Altman also addressed his 2023 comments on India’s ability to develop cost-effective LLMs, which had gone viral after DeepSeek built a generative AI model without advanced Nvidia chips and with a leaner team. He clarified that his remarks had been taken out of context, emphasising recent progress in AI distillation.