Don't fight the wrong price war

Experts argue that China's aggressive AI pricing could rewrite the global playbook, while India could emerge as a cost-efficient hub for AI infrastructure. But fundamentally there is one important caveat. China is indeed changing the AI game. But India should be extremely careful about copying the Chinese price tag and calling it strategy. Just because India has spent decades winning the global race for 'cheaper', it cannot win this race. IT services? Cheaper. Back-office processing? Cheaper. Call centres? Cheaper. Software engineers? Cheaper. And now we want to become the discount aisle of AI. That is not a sovereign AI strategy. China's AI pricing is far more consequential than a simple price war. Chinese companies are using cost as a weapon of technological diffusion. DeepSeek's latest V4 Pro initially cost just $0.87 per million output tokens, although it has since raised peak-hour pricing to $3.96. Even after the increase, it remains dramatically cheaper than many US frontier models. Alibaba's Qwen models are similarly aggressive. The result is that AI is becoming abundant, not merely cheaper. And abundance changes markets. China learnt this lesson in solar panels, batteries, electric vehicles, telecom equipment, and consumer electronics. First reduce the cost. Then increase adoption. Then build scale, and then squeeze competitors. Finally, dominate the ecosystem. Reliance Jio seems to have caught on the same strategy, if you have to draw a parallel. AI may be following the same choreography. But here is where the Indian opportunity gets misunderstood. India should not aspire to be China's cheaper cousin. It should aspire to be the place where AI creates the greatest economic value. There is a gigantic difference. The cost of compute is only one component of AI economics. Power, cooling, chips, networking, data, talent, model engineering, cybersecurity, latency, regulatory compliance, and, most importantly, commercial applications determine the real economics. A `65-an-hour GPU is wonderful. A `65-an-hour GPU producing a `60-an-hour business problem is less wonderful. India is actually making impressive progress on compute. Government data says over 38,000 GPUs have been onboarded through 14 providers, with average subsidised access of about Rs 65 an hour, roughly one-third of the global average. Data-centre capacity has also risen from roughly 375 Mw in 2020 to around 1,500 Mw in 2025. So yes, the infrastructure story is real, but AI is becoming a power industry disguised as a software industry. ALSO READ Inside Track by Coomi Kapoor: Outsiders unwelcome Data centres already consume enormous quantities of electricity, and AI workloads are pushing rack densities and cooling requirements sharply upwards. Industry estimates put power at roughly 40-60% of data-centre operating costs. Although startups like Rezonent Microchips is aiming to reduce this consumption, India cannot simply announce another thousand data centres and declare victory. The country will need reliable electricity, transmission infrastructure, renewable power, storage, cooling, and water. Otherwise our AI strategy will eventually encounter an ancient Indian technology: the power cut. And there is an even bigger danger in the 'India as a low-cost AI infrastructure hub' thesis. We could become fabulously successful at hosting everybody else's intelligence while remaining mediocre at owning our own. That would be the AI equivalent of the old outsourcing trap: India writes the code, someone else owns the platform, captures the IP, controls the customer, and pockets the disproportionate economics. The IndiaAI Mission is at least attempting to avoid that fate. It has a `10,370-crore outlay and explicitly targets compute, indigenous foundation models, data sets, startup financing, and AI talent. That is the right direction. But the ambition must go further. India needs to make some hard decisions. Why should our competitive advantage be cheap AI when our real advantage could be uniquely useful AI? India has something China and the US cannot replicate easily: the largest population-scale laboratory of digital public infrastructure, linguistic diversity, frugal innovation, and heterogeneous consumers. Had India looked at offering the cheapest payment infrastructure, UPI would not have become transformative. It became transformative because India built infrastructure that millions could actually use. AI needs the same philosophy. Imagine Indian AI becoming exceptionally good at healthcare triage, agricultural advice, vernacular education, legal assistance, logistics, manufacturing, financial inclusion, and government services. That is where the real prize lies. What's the point of following 'How cheaply can India run a GPU?' Instead, follow 'How much GDP can one GPU create?' Isn't it a radically different metric. China's price offensive could be wonderful news for India. If Chinese models force American frontier-model companies to slash prices, AI becomes cheaper globally. Recent competition has already triggered major price reductions among US providers. India should welcome this. Let China and America fight the AI price war. India should fight the AI productivity war. There is another lesson hidden in China's strategy. Cheap models encourage experimentation. Open-weight models reduce vendor lock-in. Lower inference costs allow thousands of startups to build applications that previously made no economic sense. Chinese models are gaining international traction precisely because they combine affordability, openness, and increasingly competitive performance. India should exploit that opening aggressively. But please, let us retire the phrase 'cost advantage' as our national technological destiny. Cost advantage gets you the contract. Capability advantage gets you the customer. And ownership advantage gets you the future. India has already been the world's low-cost technology back office. The opportunity now is to become the world's high-impact AI laboratory. China may be teaching the world how to make intelligence cheaper. America may continue teaching it how to make intelligence smarter. India should teach it how to make intelligence useful at population scale. That would be a far more interesting playbook. And, mercifully, one in which we don't have to compete by offering the cheapest brain in the room. Disclaimer: The views expressed are the author's own and do not reflect the official policy or position of Financial Express.