India May Face Challenges In Pulling Off A DeepSeek Moment Just Yet

Even though India is aiming to develop a low cost AI model, in line with DeepSeek, within a year, the tech industry feels it warrants more time on account of data availability and infrastructure, and wonders if DeepSeek’s methodology can be replicated easily.

With DeepSeek becoming a major player in AI foundation models, China has achieved notable advancements in the field of artificial intelligence. The new platform garnered praise from big tech giants. Even US President Donald Trump said that China’s model is 'positive' and a wake-up call for the US industry.

“The excitement, euphoria, and hype surrounding one company's success doesn't imply that others can easily replicate it. DeepSeek developed its technology with lower costs and fewer resources. While there’s nothing stopping Indian startups from doing the same, whether they can succeed remains to be seen,” said Jaspreet Bindra, managing director and founder of advisory firm Tech Whisperer.

India is now taking a more proactive approach. Union IT Minister Ashwini Vaishnaw on Thursday said that the government has created a framework calling for proposals to develop the country's own foundational AI model. He also highlighted that there are at least six developers, startups, teams that will be building a foundational model within 4-10 months.

Sandeep Dutta, chief practice officer – Asia-Pacific & Middle East region, for AI company Fractal, underscored the need for the government to enable access to data. “I don't think we fully also understand the kind of support DeepSeek might have had from the ecosystem on getting this done,” he said.

The government plays a crucial role in enabling access to data, often behind firewalls, which the industry can leverage for innovation. While capital isn't a major issue—as Indian companies can secure funding— challenges remain with data access and hardware, especially with the scarcity of chips. These factors need to be addressed for applications to succeed, he added.

Soumendra Mohanty, chief strategy and innovation officer, Tredence, echoed that in India, the challenge is access to quality, labeled, and annotated data. This is a significant hurdle in comparison to developed countries, where data integrity is more established, he said.

He also pointed out that the framework for responsible AI practices needs to evolve, and monetisation of the models has to be figured out, for Indian models to be a success.

“As AI tools become more affordable, there's a risk they might be used without fully understanding their data, biases, or explainability. Further, monetising these AI innovations is essential for sustainable business models, but Indian businesses remain cautious in adopting AI, as its applications may not yet meet their needs,” Mohanty said.

Industry players have also expressed growing cynicism over the claims made by DeepSeek, particularly regarding the extent to which their technology was developed at lower costs and with reduced computational requirements. Many are skeptical about the actual accuracy of these assertions, with some questioning whether the results are genuinely achievable within such constraints, calling for verification as well.

India has mainly relied on US-based foundational models to build solutions for its unique challenges. While a cheaper Chinese model now exists, security concerns over their use, driven by geopolitical movements, may hinder the adoption of DeepSeek in India.

The government has been restricting Chinese technology, as they do not fall under trusted category sources, since the border clashes with China. India has had a history of bans amid security implications, with Chinese apps like TikTok and equipment by companies such as Huawei and ZTE being restricted.

“The geopolitical context will influence the adoption of tools like DeepSeek, but we’ll need to adapt and build on them. Rather than banning such technologies outright, the likely outcome will be a regulated approach where we can learn from others, but are encouraged to develop our own solutions,” said Mohanty.

However, Bindra underscored that DeepSeek is an open-source model which allows hosting of data in different countries. Companies leveraging the model can store data in different countries, which helps avoid concerns. More data which DeepSeek avoids right now, like information on the Dalai Lama and Tiananmen Square, can be fed into the system.

For instance, Perplexity AI’s CEO, Aravind Srinivas, announced that the company has incorporated DeepSeek AI into its platform, allowing users to access the AI model privately and without censorship.

In a broader sense, Bindra noted that if the replication of DeepSeek’s methodology is as simple, in the coming months, the world will see many foundational models being built, truly democratising the space. This will result in reduced dependence on any one country’s foundational AI models.