Opinion For India, why the AI arms race is far from over
As China’s DeepSeek has shown, developing AI models does not require billions of dollars or access to cutting-edge GPUs; rather, it needs innovative ideas
Deepseek app is seen in this illustration taken on Tuesday. (Photo: Reuters) 
-
Share -
-
-
Adam Osborne, who developed the first successful portable computer in 1981, the Osborne 1, predicted that his company would reach $1 billion in sales by 1984. However, his company collapsed in less than 30 months, and Osborne Computer Corp went bankrupt. Very few remember Osborne and his contribution; instead, the credit goes to Apple’s Steve Jobs for making the personal computer mainstream. That’s the nature of tech — it evolves quickly, and it’s often hard to predict who will come out on top. A similar story played out with a small AI start-up from China, DeepSeek, which spooked big tech investors on Wall Street with its generative AI chatbot, a direct rival to ChatGPT.
The chatbot is able to perform some tasks at the same level as recently released models from OpenAI and Meta, despite claims that it costs a fraction of the money and time to develop, and doesn’t rely on expensive GPUs (graphics processing units). This has introduced a new approach to building AI models with fewer resources through reverse engineering — a technological breakthrough, as many AI experts are calling it — that fundamentally challenges how US tech companies have been approaching AI development.
The question many have is: Where does India stand in the global landscape? Well, India currently plays third fiddle in the AI race. The US and China are the two key contenders, fiercely competing for the crown. However, it’s not about the US vs China, but more about open-source versus closed AI models. India still hasn’t lost the race — AI is still in the early phases of development and the rules and legal guardrails are still unclear.
The new AI model released by DeepSeek is a lesson for India (and also a source of hope): It is possible to kick-start the AI development journey on its own terms, without investing billions of dollars into building massive data centres or flooding the process with vast amounts of data to improve results. The architectural advancements made by China demonstrate that similar, or even better, AI models can be developed with much less money. China is essentially saying that AI is not as expensive as it has been made out to be. This challenges several fundamental assumptions about AI progress — assumptions the US has been so confident in, treating its approach as the “gold standard”. In reality, China reverse-engineered the process and came up with a better solution to develop a cutting-edge AI model in-house, using an open-source approach with a team of 200 people, mostly recent graduates from local Chinese universities.
This says a lot about how China got it right — maybe not on the first attempt. After all, several other Chinese tech companies have developed AI models, but none have come close to matching the accuracy and performance of models from the likes of OpenAI. DeepSeek managed to do it, finally — something the US was confident China would never achieve. Silicon Valley was under the impression that they were two to three years ahead of their Chinese counterparts.
Some may call China’s attempt at a world-class AI model a fluke, because the US is still dominant in AI. But New Delhi needs to take this seriously and start working on a plan to focus more on research and development — something that allowed China to quickly close the gap with the US in AI progress. DeepSeek’s success shows how the emphasis on research has played a critical role in the development of an AI model comparable to US companies like OpenAI and Meta, despite being trained with fewer resources. DeepSeek founder Liang Wenfeng avoided the traditional route of China copying the US, such as building apps or making cheap smartphones. Instead, he focused on talent that could be used for research and resources on creating a model that could outperform OpenAI.
India has not yet had a golden moment in tech. We have never built global companies on the scale of Google or Apple, nor have we seen our tech go global. We have recently made some progress in fintech, with our Unified Payments Interface (UPI) becoming a global payment system — but that’s about it.
With Prime Minister Narendra Modi alongside major world leaders attending the AI Action Summit in Paris this week, much is riding on how India positions itself in the global AI landscape. For India to succeed, it needs to focus on two key areas. First, adopting a long-term AI-first approach, with an emphasis on developing AI technologies and creating tech platforms locally, while aiming to export these innovations globally. Similar to what China has done, India needs to set multiple smaller goals after identifying the sectors that require the most attention. However, achieving these goals may take years, and the right policies can accelerate progress.
In the immediate future, India’s greatest opportunity (which China does not have) lies in its geopolitics strategy, particularly in the context of artificial intelligence. India stands to benefit the most by being a strategic ally of the US, ensuring a continued semiconductor supply and kickstarting the development of AI technologies. The second step (and perhaps the most important) is to collaborate on developing AI technologies, meaning India must work closely with the US and Europe and establish joint ventures (JVs) so that its local companies have equal representation. However, while developing AI, India also needs to be a part of creating a specific framework around responsible AI development and use.
India needs to move away from the notion of being a follower in tech, a label it has held for years. The country has always had talent and institutions like the IITs, but for decades, these technical institutions have primarily served as a vehicle to place students with US tech giants. When the best minds leave India to either pursue PhDs at foreign universities or choose to work for companies like Google and Microsoft, it is the US that benefits, not India. What has India gained in return? Neither the technology nor the talent.
India needs to build AI technology from scratch — completely indigenous and not reliant on OpenAI or Google — and a lead in AI can only be achieved if the government, institutions, and domestic companies come together to invest in research, create an ecosystem, and develop the right AI infrastructure. AI is complex and requires a mix of programming, mathematics, and statistics. These skills are necessary to analyse data, develop efficient algorithms, and implement AI models. It’s highly technical, and developing AI models requires smart people with the right technical skills, patience, and a long-term vision.
Maybe it’s time for India Inc. to come together, create AI research labs spread across the country, and operate them independently. They should hire PhD-level students from IITs and other top technical institutions, pay them well so that they stay in India, and let them develop and build foundational AI models while working on future AI technologies beyond LLMs. India needs to realise that any nation that manages to scale up AI, democratise it, and generate developer interest — so they can reproduce and modify model weights and methods — is the one that will control the narrative and lead the AI arms race. Not to mention, LLMs are a sort of an operating system — similar to Apple’s iOS and Google’s Android — on which developers can build new applications. With China aspiring to become the global leader in open-source AI, and the US’s best models being closed-source, the former has a better chance to expand AI at a much lower cost. But in tech, anything is possible, and India could become a contender too if it plays its cards right.
anuj.bhatia@indianexpress.com
-
Share
-
-
-
Share
