On January 20, as the world tuned into the inauguration of Donald Trump as the 47th US President, DeepSeek, a Chinese company, established in 2023 by Liang Wenfeng, a young engineer and entrepreneur, launched its first generative AI large language model (LLM) — DeepSeek R1. In a week’s time, this model became the most downloaded app in the US, sending stock markets crashing across the globe and triggering debates if the generative AI bubble would follow the dotcom bubble.
This disruptive model’s pace and capability at a fraction of the infrastructure cost has challenged the US’s supremacy, premised on its massive high-cost data centre ecosystems and high-performing semiconductor chips despite China’s dominance in rare earth metals and engineering capability. As the top leaders of the world gather in Paris on February 10-11 to deliberate on “action AI”, this shock prompted by DeepSeek is going to feature in discussions; it will also give direction to how the race for generative AI will have to be handled.
Generative AI is not just a technological frontier; it is a geopolitical battleground. The ability to develop and control foundational models confers significant economic and strategic advantages. Countries that lead in AI innovation can shape global standards, influence international norms, and gain a competitive edge in industries ranging from defence to healthcare. The US-China rivalry in AI is a case in point. The US has long been the dominant force in AI research and development, thanks to its world-class universities, tech giants, and venture-capital ecosystem. However, China’s concerted efforts to become a global AI leader by 2030 have narrowed the gap. DeepSeek’s success is emblematic of China’s broader strategy to achieve self-reliance in critical technologies and reduce its dependence on Western imports.
For other nations, this rivalry presents both challenges and opportunities. On the one hand, the concentration of AI expertise and resources in a few countries risks creating a new form of technological colonialism, where smaller nations become dependent on foreign AI systems. On the other hand, the democratisation of AI tools and open-source initiatives offers a pathway for countries to build their own capabilities and assert their sovereignty in the digital age.
India, with its thriving tech industry and vast pool of engineering talent, is uniquely positioned to become a major player in the generative AI space. However, despite its strengths, India lags behind in developing foundational models. Most Indian AI startups and enterprises rely on foreign LLMs, which limit their ability to innovate and cater to local needs. The effort to purchase 10,000 graphic processing units (GPU) is yet to gather steam to be able to support the modest AI projects being planned around optimal healthcare delivery, improved crop yields, and enhanced access to personalised education.
Investing in foundational models is not just a matter of national pride; it is an economic imperative. Generative AI has the potential to add trillions of dollars to the global economy, and India cannot afford to be a passive consumer in this transformative wave. Moreover, India’s diverse linguistic landscape presents a compelling use case for localised AI models. With over 22 officially recognised languages and hundreds of dialects, India requires AI systems that can understand and generate content in multiple languages. Foundational models tailored to Indian languages would not only address domestic needs but also create export opportunities in other multilingual regions.
To achieve this vision, India must adopt a multi-pronged strategy. First, the government should increase funding for AI research and development, particularly in academia and public-private partnerships. Initiatives like the National AI Strategy and the establishment of AI research centres are steps in the right direction, but more ambitious efforts are needed. Second, India must invest in infrastructure to support AI innovation. This includes high-performance computing resources, cloud infrastructure, and data-sharing frameworks that enable researchers and startups to train large-scale models. Collaboration with global tech leaders and participation in international AI consortia is crucial to accelerate progress. Third, India must foster a culture of innovation and entrepreneurship in AI. This requires not only technical skills but also a mindset that embraces experimentation and risk-taking. Initiatives like AI hackathons, startup incubators, and industry-academia collaborations can nurture the next generation of AI leaders. Fourth, Indian tech giants should start the LLM pursuit as India has access to better GPUs and interconnects compared to China. The iCET sets India on the path of comprehensive tech cooperation with the US.
DeepSeek’s rise is a testament to the transformative power of generative AI and the shifting dynamics in the global AI landscape although it still remains to be seen what is the fairness in all its responses, particularly when it comes to prompts about China and its political systems. It has stated that its R1 model is open source and has laid out its reasoning in a paper in Github. So, there is the scope to test and validate that model and see if such models can be repeated at reasonable costs, factoring in concerns about bias, privacy, and misuse.
The writer, a defence and cyber security analyst, is former country head of General Dynamics