India’s AI ambition: Costs beyond code

India wants a seat at the front of the global AI table. That promise includes new industries, higher productivity, and better public services. It also forces hard choices: scarce public funds, social dislocation, and sharper demands on electricity and water. Pursuing leadership without clear priorities will concentrate gains in a few hubs, pull money from basic services, and stress communities already under strain.

Building large-scale AI is costly. Compute clusters, specialised chips and fabrication plants require heavy capital. Training big models needs continuous power and cooling. Chip fabs demand ultra-pure water and steady utilities. For a country still expanding schools, clinics and rural electrification, each rupee spent on hyperscaler-scale infrastructure is a rupee not available for basic services.

More from Opinion

India’s advantages matter: deep engineering talent, an established IT services industry, and a digitally active population. These strengths lower some barriers. They do not remove structural trade-offs. When computers and fabs cluster in a few cities, jobs, contracts and intellectual property concentrate there. Smaller towns and rural districts risk being left behind while shouldering environmental burdens from nearby facilities.

Environmental effects are immediate. Many regions already face seasonal power shortages and water stress. Data centres and fabs amplify those pressures. Cooling systems and process water withdraw from local basins and compete with farms and households. Electricity meters record obvious draws; planning often ignores the equally crucial need for treated water and reliable grid support. Expanding AI infrastructure without strict efficiency and local safeguards will intensify conflicts over resources.

The social impact is equally clear. High-value AI jobs settle in metropolitan hubs. Without targeted skilling and geographically distributed opportunities, migration will rise and urban services will strain. Models trained on global datasets often miss India’s linguistic and cultural diversity. The result: technologies that work for some Indians and exclude many others.

Fiscal prudence requires prioritisation. Policymakers should weigh the cost of national-scale compute and fabs against alternative public investments. Rather than racing to host the biggest data centre or the most advanced chip plant, India should target investments that deliver measurable public value. Priorities should include shared regional compute, research that addresses national needs, and pilots that prove returns in health, agriculture and governance.

Practical steps drawn from lessons learned over the past few years:

Publish environmental metrics. Every large training run, data centre and fabrication plant should report standardised measures of electricity and water use. Public data creates accountability and enables realistic efficiency targets.

Build shared regional compute hubs. Invest in public or public–private cloud and edge facilities designed for local needs. Shared infrastructure lowers costs for startups, universities and government projects and cuts redundant energy and water use.

Tie incentives to efficiency and renewables. Offer tax breaks, land concessions and subsidies only to projects that meet low PUE (Power Usage Effectiveness) targets, implement aggressive water recycling and deploy on-site renewables. Promote closed-loop cooling and wastewater reuse and favour sites with sustainable water plans.

Fund local languages and datasets. Create open corpora and support research so models reflect India’s linguistic and cultural plurality. Public datasets reduce reliance on foreign models that may not match local priorities. Homegrown Sarvam is a good start.

Scale skilling where people already live. Set up regional centres of excellence and certify apprenticeship and upskilling programmes outside major metros. Local training widens talent pipelines and eases migration pressure.

Treat digital infrastructure as development finance. Development banks and domestic financiers should include compute access, reliable power and water stewardship in economic policy. Loans and grants should favour sustainable deployments that align with other development goals.

Phase expansion with social safeguards. Where AI threatens jobs, require transition programmes and incentives for firms to hire locally.

Use procurement to shape markets. Government buying can favour energy-efficient solutions, open-source models tailored for India, and vendors that commit to local reinvestment and skills development.

Without guardrails, ambition repeats familiar mistakes: pollution and resource stress concentrated in vulnerable places; public funds diverted to prestige projects; and technical leadership captured by a few private actors. India can avoid that pattern by treating compute and fabrication as means, not ends.

When investments target clear public goals such as better rural healthcare, language-aware education, climate-smart agriculture, and local industry modernisation, they become easier to justify and regulate.

India need not choose between leadership and responsibility. Leadership in AI must be measured: transparent metrics, phased investments, firm efficiency standards and social policies that spread opportunity. Align the AI push with budget and environmental constraints, and India can build a model of technological leadership that is powerful, fair and sustainable.

(Sreejith Sreedharan is a technology analyst and author. The views expressed in this piece are personal and solely those of the author. They do not necessarily reflect Firstpost’s views.)