China’s development of its foundational artificial intelligence (AI) model, DeepSeek, has inadvertently settled a crucial debate in India: whether to invest in large language models (LLMs) or focus on small language models (SLMs). Some tech industry leaders have argued that SLMs should be the priority, given their commercial viability and India’s expertise in service-oriented AI applications. However, with China successfully launching DeepSeek at a lower cost than Western alternatives, India can no longer afford to sideline foundational AI model development.
The Indian technology ecosystem has, in the past, attempted to build alternatives to American giants’ products. Messaging app Hike was built on the lines of WhatsApp and microblogging platform Koo as an alternative to Twitter. But they struggled against their global counterparts and wound up. Comparisons with China are misplaced. China is a closed market that blocks platforms like Google, Facebook, and Twitter, allowing domestic firms like WeChat and Weibo to thrive. India is an open economy where US tech products enjoy wide popularity, alongside Chinese consumer goods. So any AI foundational model built in India will face competition from American and Chinese alternatives.
This challenge is not hypothetical. India’s 4G telecom stack, developed by TCS and Tejas Networks, has not found domestic takers beyond BSNL. The competition in AI will be even tougher, requiring India’s foundational model to prove its superiority — also in terms of cost — over global players. The question is not if India should develop its own foundational model — it must, as every nation should strive for technological sovereignty — but whether it can create a sustainable market for it.
OpenAI CEO Sam Altman, during his recent visit to the country, highlighted India’s growing significance in AI. He acknowledged India as OpenAI’s second-largest market, with the number of users tripling in 2024. This underscores strong AI adoption, but poses a paradox: users are enthusiastic about tech products, but they overwhelmingly rely on foreign models rather than indigenous ones.
The emergence of DeepSeek and Altman’s comments have reignited discussions on India’s AI strategy. While the Centre is investing in infrastructure, with a Rs 10,000-crore IndiaAI mission aimed at building indigenous foundational models, the real test lies in execution. The plan includes developing AI models that cater to India’s linguistic diversity and cultural nuances, supported by an advanced computing infrastructure powered by 18,693 graphics processing units. Its success will depend on whether these models can compete with OpenAI, Google’s Gemini, or DeepSeek’s offerings in performance, efficiency, and usability.
One of the main reasons why India’s AI ambitions face an uphill battle is that Indian IT companies primarily focus on software services and enterprise solutions, while firms like OpenAI and Google invest heavily in core AI research and product development. Expecting them to develop cutting-edge LLMs without a viable revenue model would be unrealistic.
At the same time, India’s AI start-ups are exploring alternative pathways. Many are considering switching to open-source AI models such as DeepSeek and Meta’s Llama, which allow them to build customised solutions without the prohibitive costs of training foundational models from scratch. The rise of open-source models has led to declining API costs, making AI more accessible to start-ups. As AI continues to evolve, affordability and differentiation will be key to market success.
The government is making strategic moves to address these challenges. The recently announced AI safety institution will focus on mitigating bias, enhancing privacy, and ensuring responsible AI deployment. The IndiaAI mission would fund projects that leverage AI for social impact in areas like agriculture, climate change, and education. These align with India’s broader digital public infrastructure strategy, which has yielded success with Aadhaar and the United Payments Interface.
India must adopt a dual approach: develop foundational AI models while fostering a strong ecosystem for applied AI solutions. OpenAI’s growing footprint in India shows there is significant demand for AI-powered applications. Indian start-ups must leverage this to build innovative products, whether by using OpenAI’s models or developing their own.
In the long term, India’s AI aspirations will depend on three key factors: access to computing power, talent, and market adoption. While the government is addressing the first two through infra investments and research funding, the third remains the biggest challenge. Creating a successful AI model is not just about technical prowess; it requires a well-defined go-to-market strategy, a supportive regulatory environment, and a user base willing to adopt indigenous solutions.
For now, the government seems to have come to view AI not just as a tech pursuit but also an economic and strategic imperative. The rise of DeepSeek has proven cost-effective AI models are feasible, even in a highly competitive landscape. While it’s good that India has taken inspiration from this, it must recognise its unique market dynamics. Creating a foundational AI model is one part, but carving out a sustainable niche in the global AI ecosystem is where the real challenge lies. While me-too products mushroom when something is trendy, they are fast to fade when faced with hurdles related to funding and market adoption. This is a bigger task than making a domestic foundational model.
