India should avoid attempting to compete with China's aggressive AI pricing strategy by focusing instead on creating economic value through useful applications. While China uses cost as a tool for technological diffusion—seen in DeepSeek's V4 Pro, which initially cost $0.87 per million output tokens—India risks falling into an outsourcing trap if it merely hosts intelligence rather than owning the intellectual property.

The Indian government is making progress in compute infrastructure. According to government data, over 38,000 GPUs have been onboarded through 14 providers, with average subsidised access costing approximately Rs 65 an hour, roughly one-third of the global average. Data-centre capacity has also grown from roughly 375 Mw in 2020 to around 1,500 Mw in 2025.

However, experts warn that infrastructure alone is insufficient. AI economics are determined by several factors, including:

• Power and cooling
• Chips and networking
• Data and talent
• Model engineering and cybersecurity
• Latency and regulatory compliance
• Commercial applications

Industry estimates suggest power accounts for roughly 40-60% of data-centre operating costs. To succeed, India requires reliable electricity, transmission infrastructure, renewable power, storage, cooling, and water.

The IndiaAI Mission has a ₹10,370-crore outlay targeting compute, indigenous foundation models, data sets, startup financing, and AI talent. Rather than competing on the cost of running GPUs, the strategy should aim to maximise how much GDP a single GPU can create. Potential high-impact areas include healthcare triage, agricultural advice, vernacular education, legal assistance, logistics, manufacturing, financial inclusion, and government services.