The rapid rise of China in AI signals a significant shift in the tech landscape. The long-held assumptions about Western dominance no longer hold true, raising questions about the balance of power in the AI industry as well as raising possibilities for nascent powers like India.

Now, the global AI competition is no longer just about innovation. It's about who can set the pace for the future of this very important technology.

As India gazes into this chasm of chance and consequence, to hesitate now would mean losing the future to others who wouldn't.

Foundation AI And Its Role In Global Competitiveness 

The US, with its glittering Silicon Valley, Ivy League talents and Pentagon's calculated strategies, has poured a cumulative US$ 249 billion in AI from 2013 to 2022. Following his return to the White House, President Donald Trump announced the Stargate Project, a collaboration with OpenAI, Oracle and SoftBank.

The project aims to invest upto US$ 500 billion in AI infrastructure over the next four years, starting with a US$ 100 billion data centre in Texas. It is expected to create over 1,00,000 jobs and further solidify the US’ leadership in AI development.

Meanwhile, necessity has become the mother of invention in China. The country is making significant strides in the AI world, emerging as a cost-efficient powerhouse.

DeepSeek, the foundation AI model, has been made at less than US$ 6 million, that too without cutting-edge chips like the NVIDIA A100, which have powered the AI revolution brought on by ChatGPT in 2022. DeepSeek's models — namely V3 and R1 — were built at a fraction of the cost of its Western equivalents and have taken approximately two months to create from scratch.

This has had a ripple effect in the US and European stock markets, triggering a US$1 trillion selloff, as investors have begun questioning the high valuations of major American tech firms. Share prices of companies like Nvidia, AMD, Google and Meta took a nosedive on Monday. The Chinese model also took the top-of-the-charts position on Apple's app store.

This is astonishing because the US had placed crippling sanctions on China, restricting the flow of the most advanced AI chips from Nvidia and AMD into the hands of the dragon nation. 

But China continues to do what it does best — creating something as good that is also cheaper. And its great feat raises several questions. For India, and for the world.

If a small lab like DeepSeek can do this, what’s stopping India's AI ecosystem from leapfrogging? Does it even make sense to invest so much in AI infrastructure, when better results can be achieved using less GPUs?

Open-Sourced Vs Proprietary

Unlike the traditional AI models that are designed for specific functions, foundation AI models serve as versatile platforms upon which specialised applications can be built. Hence, the development of proprietary foundation models is crucial, as it ensures control over the underlying technology that allows for customisation to meet specific needs.

Foundation models like DeepSeek are important for another reason. They are open-sourced — meaning, anyone can access, modify and build upon them to foster innovation and transparency in the field.

This contrasts with proprietary models, like the ones released by OpenAI and Google, which are often locked behind paywalls, thus limiting who can use or improve upon them. The open-sourced nature of China's models allow researchers, developers and organisations worldwide to contribute to, and benefit from their development.

India’s geopolitical challenges and vast geography make foundation AI indispensable for her security, given the need to surveil contested borders with not-so-friendly neighbours. Cyber-defence systems could leverage AI models to neutralise state-sponsored hacking attempts, like the 2022 breach of AIIMS servers. 

But the spectre of dependence looms heavy on India. Relying on foreign AI models is inviting vulnerabilities at the heart of critical systems. In the past, India’s reliance on Russia for cryogenic engines for space exploration exposed the precariousness of foreign technology. History warns against such reliance, with technological dependence turning into strategic subjugation.

For India, open-sourced models can be particularly transformative.

They lower entry barriers for startups, academic institutions and government projects that may not have the financial or technical resources to develop propreitary models from scratch. By leveraging open-sourced technology, India can build solutions tailored to its unique challenges, such as regional language processing, agriculture or public healthcare systems.

This gap in resources and talent underscores how far India needs to travel to bridge the divide in AI development.

India's supercomputers pale in comparison to those with the US and China. In the field of academia and innovation, the two powershouses own more than 60 per cent of the world's AI talent.

The foreign datasets, which form the bedrock of AI models, also carry their own biases, infusing the models with cultural and strategic ideologies of their creators. Therefore, India’s absence in this field is a chasm.

AI models are already becoming the backbone of military, economic and social power in the 21st century.

And, the AI race may not become a zero-sum game, where the gain of one country or company would mean the equivalent loss for another. Victors alone won't shape the norms, values and infrastructure of the global digital age.

Rather, many nations can be superpowers in their own right. But without its own foundation AI model, India risks falling into the shadow of dependence; and its sovereignty getting tethered to the innovations of others.

A Strategic Framework For Indian AI Development

The path forward demands boldness with an all-round strategy that would address both technological and strategic imperatives. India must boost AI R&D funding, aiming for around 3 per cent of GDP by 2030, akin to China's 2017 strategy.

Partnerships with countries like Japan could help innovate exascale supercomputers. India’s ocean of untapped talent in its demographic dividend must be harnessed. Engineers should be trained through initiatives like the NASSCOM Future Skills Prime.

Military-civil fusion and the weaving of startups into defence R&D hold immense promise, with enterprises like Grene Robotics, ideaForge and Big Bang Boom Solutions poised to play important roles. Beyond security, AI can be seamlessly integrated into fields like monitoring crop patterns, personalised medicine and emergency disaster response.

(The writer is a researcher on security and society at the Advanced Study Institute of Asia, SGT University. Views are personal)