India must ride the next AI wave, not chase past ones

The next revolution will come from smaller models, specialised models, reasoning systems, neurosymbolic AI, edge AI, agentic workflows, and systems that can verify their own output instead of producing fluent nonsense

- Vivek Wadhwa

While India debates — duly impassioned by China’s recent AI strides — spending billions on building its own foundation model, a company that has already spent billions on the world’s most famous AI laboratory is now looking at using Chinese models.

That company is Microsoft and its CEO, Satya Nadella, has said that the world cannot allow a handful of AI giants to eat the economy. Microsoft has invested billions in OpenAI, helped turn ChatGPT into a household name, and made AI the centre of its future. Yet, he recently told the Wall Street Journal that he is considering using technology from DeepSeek, the Chinese AI company that shocked Silicon Valley by showing that powerful models could be built and run far more cheaply than anyone believed. Indeed, China’s Z.ai recently revealed capacity that rivals Anthropic’s frontier models in terms of cybersecurity implications.

Nadella’s intuitiveness should sober up India’s AI chest-thumpers.

If Microsoft, with all its money, cloud infrastructure, engineers, enterprise customers, and inside access to OpenAI, is preparing to use cheaper Chinese models, why should India spend billions trying to build a giant model of its own?

Microsoft has seen the future, and it does not look like billion-dollar models with trillion-dollar egos. It looks like cheap, interchangeable intelligence running inside platforms that own the data, security, workflows, compliance, pricing, and customer relationship. The model builders may have the glamour; the platform owners will collect the rent.