India should be a leader in building small AI models, says Sam Altman on second visit

On his second visit to India in two years, OpenAI CEO Sam Altman reaffirmed the country’s potential as a frontrunner in artificial intelligence, particularly in building small and reasoning models. Amid intensifying AI competition from Google’s Gemini, DeepSeek, and others, and increasing global scrutiny over AI’s impact, Altman’s India tour underscores the country’s growing significance in the AI ecosystem.

“India should be a leader in building small models, especially reasoning models,” Altman said, emphasising that while AI training costs will continue to rise exponentially, the returns in intelligence and revenue will also grow significantly. According to him, near-term AI models are already reaching the threshold of being good enough to address critical issues like healthcare and education—sectors where India has much to gain from AI-driven innovation.

India is now OpenAI’s second-largest market, highlighting the country’s fast-growing adoption of AI-powered tools. Altman encouraged India to “do everything within the AI stack,” indicating that beyond just using AI, the country should actively build and contribute across the AI value chain.

“It’s amazing to see what India has done so far,” he added, acknowledging the strides Indian startups, researchers, and developers have made in AI innovation.

Altman also addressed the rising costs of AI model training, saying that while it’s still expensive, the cost per unit of AI intelligence is falling by a factor of 10 every year. However, he pushed back against the notion that this will reduce the need for AI hardware, implying that the demand for AI infrastructure will continue to grow.

Additionally, he clarified that his past comments on building foundational models were taken out of context, perhaps referring to previous debates on whether India should focus on developing its own large-scale AI models or leverage existing ones.

Delhi marked the last stop of Altman’s Asia-Pacific tour, during which he engaged with policymakers, AI researchers, and business leaders. His visit comes at a time when India is ramping up its AI ambitions, with initiatives like Bhashini (for AI-driven language translation), government-backed AI compute infrastructure, and a growing focus on AI regulation.

While OpenAI is still working on AI distillation (making models smaller and more efficient), Altman acknowledged that progress in this area has not been “incredible” yet. His remarks signal that small, cost-effective models might be a key area where India can take the lead—a sentiment that aligns with the country’s AI goals.