AI is the new arms race. India must catch up.

Countries are competing over Artificial Intelligence as though it were the new global arms race. This is not surprising if you look at what AI can do for nations. AI capabilities can boost the economies of nations. The only debate is by how much — some think it could be like a new Industrial Revolution.

From finance to manufacturing, there will be no aspect of the economy that will not be touched by AI. Automation will boost productivity and hence growth.

It will also revolutionise warfare: autonomous drones, faster intelligence processing and advanced cyber warfare will give countries with superior AI capabilities a strategic edge. It can even be a deterrent like nuclear weapons.

Given these advantages, governments also need to worry about tech sovereignty. Depending upon other countries’ AI data centres and software makes one vulnerable.

AI depends on big data and countries will also want control over the data itself. Given that the global AI revolution is still in its infancy, the global conversation on regulation and cooperation will privilege countries with advanced AI capabilities.

China shows the way

Realising the strategic power of AI, the United States banned the sale of advanced AI chips of the sort made by Nvidia, to China, in much the same way that superpower countries decide who they want to sell their missiles and fighter jets to.

Necessity is the mother of invention. China’s DeepSeek released a new R1 model that, it turns out, is just as capable as OpenAI’s first O1 model.

Large language models require immense computing power and expensive chips. Moreover, China achieved this feat despite being denied the AI chips made by Nvidia.

The US is by no means going to be left behind in the AI race. It has announced an ambitious Stargate project worth $500 billion which will be a joint collaboration between OpenAI, SoftBank, Oracle, Microsoft and Nvidia.

Not a bipolar race

For now the global AI race looks like a contest between US and China but DeepSeek’s achievement has changed everything for the rest of us.

On 10-11 February, India will co-chair the Paris AI summit. The summit seeks to shift the European conversation on AI from an obsessive focus on regulation to innovation.

The feeling in Europe is that it is getting left out of the AI revolution. That is also an immediate realisation in India after DeepSeek suddenly bursting onto the scene.

India has not been sleeping on the AI front. There has been a lot of conversation in both the IT industry and government. But since India’s IT industry is focused on services, the consensus in India was that we should let the US and others build the expensive AI chips and models and we will happily use them.

Indians were early and eager adopters of ChatGPT but the need for indigenous AI capabilities for reasons of national security, sovereignty and economic advancement dictate that India must build its own foundational models.

Jugaad AI

In September 2023, OpenAI founder Sam Altman was asked at a summit in India whether it was possible for an Indian start-up to make a foundational new model for around $10 million. He said it would be a hopeless enterprise but you must try anyway.

He has been proved wrong not by an Indian start-up but a Chinese one.

Sam Altman and others have wondered why Indians have been such enthusiastic users of ChatGPT. Indians are the second largest users of ChatGPT after the US. This is perhaps because generative AI gives us shortcuts to research and execution and Indians love shortcuts — the famous Indian “jugaad”.

There’s no reason why an Indian company cannot build a low-cost and yet world-class large language model, just as we can send missions to the moon at the cost of a Hollywood movie. What has been lacking is ambition. DeepSeek changes that.

Make AI for the world

BharatGPT, an Indian LLM supported by Reliance Industries, focuses on Indian languages. It is nowhere near ChatGPT in its capabilities for now, having just about launched.

BharatGPT is built on Google Cloud. Like DeepSeek and others in China, we must seek to build our AI capabilities on our own computing power and data centres (Reliance and others are beginning to invest in this). Cost calculations for such Indian attempts will now change.

The Indian government will have to invest a lot more in AI research and producing AI talent, not least because Indian AI talent will be quickly taken away by the US and others.

The government must also encourage Indian AI start-ups to think of not just building applications but also foundational research and LLMs. Efforts to create an indigenous AI infrastructure is key to increasing AI adoption by companies and government institutions.

The Indian government’s mantra, ‘Make AI in India, Make AI work for India’ comes from a focus on data security. But DeepSeek has shown us why it is both important and necessary for India to make AI for the world.