Can India strike GenAI gold?

ABROAD AT HOME

Few technologies have gripped the public imagination – or investors’ wallets – like Generative AI has in recent years. The spark ignited with OpenAI’s ChatGPT in late 2022, propelling the company to a $300 billion valuation in its latest funding round. Meanwhile, xAI, Elon Musk’s venture leveraging X data to power tools like Grok, boasts an $80-billion valuation. This surge prompts four key questions: What exactly is generative AI? Where’s the money coming from? What’s the strategic landscape? And what role does India play in this unfolding story?

Predictive AI, often synonymous with traditional AI/ML (Artificial Intelligence/Machine Learning), analyses historical data – typically numerical – to forecast trends or outcomes. Think customer behaviour, market shifts, or equipment breakdowns, all powered by statistical models and algorithms.

Generative AI, by contrast, creates anew. It generates text, images, music, or even code, drawing on patterns in vast datasets. (Full disclosure: Parts of this column were rewritten by an AI). While innovative, it’s hardly flawless – hallucinations (errors) and biases from training data, like that used in models such as GPT-4, Google Gemini, Grok, Llama, Claude, or DeepSeek, can skew results. Fundamentally, it’s statistical, not deterministic, and predictable.

Interestingly enough, all this may not lead to great success for AI companies (even though Generative AI might become as ubiquitous as email). History offers a parallel: the 1849 Gold Rush in California. Miners chased riches, but most scraped by while merchants selling picks, shovels, and jeans cashed in.

Today, generative AI follows a similar script. Companies building AI solutions – custom chatbots, content tools, or industry-specific platforms – aren’t raking in big profits. Instead, the real winners are the intermediaries: chipmakers like Nvidia, whose hardware powers model training, and cloud giants like Amazon AWS, Microsoft Azure, and Google Cloud, providing the computing, storage, and networking backbone. Their indispensability ensures steady, less speculative revenue, while the risks are taken by the AI vendors.

Strategically speaking, Microsoft’s foresight has been pivotal, turning its OpenAI partnership into a market-defining move. It’s the only company to hold a top-10 market cap spot in both 2001 and 2025, outpacing Google, an early AI leader now playing catch-up.

Meanwhile, China’s entry – via models like DeepSeek – has shaken things up by 2025. Eschewing the US’ billion-dollar proprietary approach, Chinese players lean on open-source innovation, sparking competition, disruption, and debates over intellectual property.

India can benefit from these market moves if there is a proper plan of action. It stands at a crossroads with unique opportunities, and threats, in Generative AI:

Skill development and reskilling: Building expertise in this domain. Especially as agentic AI and even ‘vibe-coding’ (software generated by AI) begin to accelerate, one of India’s major current competencies in software services may be affected. Similarly, other business process outsourcing may also be affected unless there is reskilling.

Indic Knowledge Systems (IKS): Integrating India’s cultural and intellectual heritage into AI models. Given the overwhelming nature of Western content in the training data, IKS may be entirely eliminated from generative AI output. A potential solution is Indian LLMs trained on narrow domains, eg. a Panini AI or a Patanjali AI. These can then be used for research, as well as to preserve and enhance IKS. India must worry about the possible expropriation of IKS by foreign LLMs, because they will rapidly run out of data and will have to depend on “synthetic data” for training, that is data generated by LLMs. Because of peculiarities with the statistical model, synthetic data will lead rapidly to model collapse (ie. the AI produces gibberish). Thus IKS would be for them a tempting source of training data.

New business models: Creating value, perhaps through real-time language translation. By doing this, communication between Indians speaking different languages immediately becomes easy, and audio content in any language around the world becomes available in your mother tongue: potentially useful for university education.

Societal Impact: Transforming education at primary, secondary, or tertiary levels. It is universally acknowledged that mother-tongue education, for instance in the hard sciences, is far superior to learning in a second language such as English.

Yet challenges loom – copyright erosion (think the current Ghibli fad) and the risk of Indian data being siphoned off for foreign models. Deepfakes – fake videos and audio indistinguishable from the original – are also beginning to appear.

India needn’t reinvent the wheel by building foundational models from scratch; open-source options abound for adaptation. The focus should be on training data – curating high-quality datasets, from IKS to medical and financial records – while safeguarding Indian data from exploitation. This approach could position India as a global player, blending innovation with cultural preservation.