One of the defining debates in artificial intelligence is whether the future belongs to open or proprietary models.
It gained renewed momentum in recent weeks with the rise of Chinese open-weight models that rival leading US frontier closed models while being developed and deployed at a fraction of the cost.
For India, the question isn’t just technology but national strategy, digital sovereignty, and economic competitiveness. So what path should it take?
The sovereignty stakes: For Sarvam co-founder Vivek Raghavan, the big issue isn’t whether models are open or proprietary but whether India develops capabilities across the AI stack.
“If India doesn’t create a full-stack AI alternative, then we will be a digital colony,” he told us.
The case for ‘both’: Many industry leaders, however, argue that the answer isn’t choosing one side.
“You need both,” Cognizant CEO Ravi Kumar S told us. “Closed frontier models power the advances… Open-weight models can be industrialized in a much cheaper way”
Google DeepMind’s Manish Gupta also said there is no “one-size-fits-all” approach, saying that India should leverage both open and proprietary AI models depending on the use case.
Gupta points to one of the biggest advantages of open-weight AI: permanence.
- Once model weights are released, “there is no going back. Nobody can take those weights away,” he said, making them an increasingly important building block for AI sovereignty.
At the same time, he argues that startups tackling complex AI problems should continue using frontier models such as Gemini, which are becoming increasingly cost-effective.
The growing case for open models
Last month, a group of 25 leading US technology executives, spearheaded by Nvidia CEO Jensen Huang, released a letter arguing that open models are vital for innovation and national security.
- That list quickly grew to over 150 signatories including Amazon, Google, OpenAI, and SpaceX, with Anthropic being the sole standout among prominent AI players.
The letter stated that open models expand access to the AI economy while strengthening competition, helping keep the “gains of AI broadly shared rather than concentrated in a few hands.”
Beyond the model wars: For Umakant Soni, co-founder of BHARAT1.AI and AI Foundry, however, India risks asking the wrong question altogether.
Rather than attempting to compete with OpenAI, Google, or DeepSeek in training ever-larger frontier models, Indian startups should focus on building specialised AI products that solve real enterprise problems.
- Open-weight models dramatically lower the cost of innovation by giving developers access to world-class foundation models that can be adapted for Indian languages, industries and workflows, he says.
The bottomline: India doesn’t need to choose between open and proprietary AI. Instead, its long-term advantage may lie in combining both – using frontier models where cutting-edge intelligence matters, open-weight models where affordability and customisation are critical.
- The countries that emerge strongest from the AI era may not be those with the biggest models, but those with the flexibility to deploy the best model for every job.
Was this newsletter forwarded to you? You can sign up for the AI Edge here. And don’t forget to sign up to Tech3, our daily newsletter that breaks down the biggest tech and startup stories every weekday evening.