India Must Build on Existing AI Systems

Instead, it is racing to catch up on foundational models, which is yesterday's opportunity

- Vivek Wadhwa

India is pouring its energy into catching up with the US and China on building foundational Artificial Intelligence (AI) models. Government officials are discussing billion-dollar investments in data centers and computing infrastructure. Academics are calling for sovereign large language models (LLMs). Industry groups are lobbying for more public-private collaborations in model training. But all this focus is on yesterday's problem.

The world doesn't need more foundational models. It already has dozens. Open-source alternatives like DeepSeek, LLAMA, Qwen, and Mistral are freely available—and improving at a breakneck pace. DeepSeek recently released a model that rivals GPT-4 in reasoning benchmarks. Qwen, developed by Alibaba, has become a top-tier model for multilingual tasks. India doesn't need to build its own from scratch. It can take these models and run with them.

The real opportunity isn't in recreating what already exists. It's in doing what India has always done best: Building on top of what's already there.

That's exactly how India became a superpower in IT services. It didn't invent the microchip, but it built billion-dollar firms that helped the world use microchips. It didn't create operating systems, but it became the global center for enterprise software development. India didn't pioneer cloud computing, but it gave rise to SaaS (software as a service) giants like Zoho and Freshworks. It didn't create banking infrastructure, but it built UPI—the world's most advanced payments platform.