India is aiming to become a global leader in artificial intelligence (AI). From semiconductor design partnerships to the rapid expansion of data centres, the country is building infrastructure that will power its digital future and promote AI sovereignty as a national priority.
A NITI Aayog report estimates that AI could add between 500 and 600 billion dollars to India’s GDP by 2035.
India is on the threshold of a new AI-led industrial era shaped by chips, servers, and data flows. However, the physical systems that enable AI also impose significant pressures on energy, water, land, and waste management systems.
Being a climate-vulnerable country, the question is not whether India should build an AI ecosystem, but how to build it well—without embedding environmental degradation into the foundation of its digital economy.
Manufacturing: Chips That Enable AI
India’s semiconductor plans are still in their early stages. The most recent announcement includes a plan to co-design 2nm (nano-metre) chips with the UK-based firm ARM.
This is in addition to ongoing manufacturing efforts such as the Tata PSMC fabrication project in Gujarat, which focuses on mature technologies such as 28 nm and 40 nm nodes, and is advancing quickly as India moves toward building full local manufacturing capacity.
When semiconductor manufacturing begins at scale, it will bring heavy environmental demands.
Semiconductor fabrication is one of the most water- and energy-intensive industrial processes worldwide.
It requires ultra-pure water, chemicals, and cleanrooms that operate without interruption.
In India, many sites proposed for these facilities, including those in Andhra Pradesh and Punjab, already face low per capita water availability. These new facilities could place significant pressure on local supplies.
Chip-making is also a major source of emissions. Factories still use fluorinated gases during manufacturing. These gases have very high global-warming potential. The specialised infrastructure needed to build, run, and maintain fabrication units also carries a high embedded carbon footprint due to their high energy consumption.
These risks remain ahead of us, which makes this the right moment for preventive action.
Currently, disclosure requirements for water and energy use are still minimal, and environmental standards specific to semiconductor facilities do not exist. This underscores the importance of building environmental safeguards into early AI industrial policy before these facilities scale.
Without strict efficiency and emissions standards, the sector could increase the overall carbon intensity of India’s industrial growth.
Data Centres: The Physical Backbone of AI
If chip manufacturing represents India's future ambitions, data centres represent its present realities. These are facilities where AI models are trained, stored, and deployed.
Currently, India's data centre capacity is about 1.2 gigawatts. This includes data centres used across all sectors, including banking, e-commerce, telecommunications, OTT, and online gaming.
The data centre surge is expected to be two-fold. The first is for non-AI-based use cases, such as in telecom, where smaller data centres are expanding into Tier 2 and Tier 3 cities to support rising consumption patterns.
The second and more prominent one is the growth led by AI-related workloads, which is expected to scale India’s total data centre capacity to nearly 9 gigawatts by 2030.
The government sees these investments as essential to building a trillion-dollar digital economy. But there are challenges.
Three-quarters of India’s electricity generation is from fossil fuels. Every new data centre, therefore, could lock in additional emissions unless it secures a renewable energy supply at scale.
Rapid growth in AI workloads could also sharply raise the sector’s emissions footprint.
A single AI query can consume up to ten times more power than a basic online search, and training a large language model can use over 1,000 megawatt-hours of electricity, roughly equal to the consumption by several hundred Indian households.
Water scarcity is another concern. Cooling systems in large data centres rely on water-based technology, yet over 80 percent facilities today are located in water-scarce states (refer Map 1), such as Maharashtra, Telangana, and Tamil Nadu.
In Bengaluru, data centres together already consume nearly eight million litres of water each day, even as the city faces extreme water shortages.
As new campuses emerge to support AI, competition for water between industry and communities could intensify significantly.
Land use adds another layer of complexity.
Deloitte estimates that India will need 50 million square feet of land for its data centres by 2030. In 2024, villagers in Mekaguda filed a petition against Microsoft in the Telangana High Court, alleging land encroachment and contamination of a local river.
This is just one example that shows if digital growth is not managed well, it can strain local systems and lead to tensions within communities.
Today, environmental governance for data centres remains fragmented. There is no dedicated policy framework for assessing their resource use or mandating disclosures.
Certifications for green operations are voluntary and vary significantly across firms. If this continues, the industry risks embedding unsustainable practices as it scales.
E-Waste: The Industry’s Hidden Afterlife
AI expansion produces fast hardware turnover. This makes e-waste an unavoidable part of the system.
High-performance chips and servers have short lifespans, often two to three years, before newer generations replace them.
This cycle produces large amounts of toxic electronic waste that contains rare earth metals and hazardous materials. If not properly recycled, these substances can contaminate soil and groundwater.
E-waste has already begun to overwhelm our existing management systems with a 73 percent increase in the past five years.
The country produced an estimated 4.17 million tonnes in 2022. Only one-third of this was processed through proper channels.
The current regulatory framework’s operational details largely reflect consumer and office equipment pathways and are not tailored to the management of data centre-specific e-waste.
The Extended Producer Responsibility (EPR) rules assign manufacturers the duty to collect and recycle waste, but on-the-ground enforcement is currently weak, and reporting remains limited.
Without updated standards for AI-related hardware and a substantial collection and recycling system, the environmental impacts of e-waste can outlast the technology itself.
These consequences often fall hardest on vulnerable communities. Without reform, the ecological footprint of AI will remain hidden but severe.
Emerging Governance Gaps
A clear pattern runs through the above-outlined sectors. India’s environmental governance was built for traditional industries and has not yet adapted to the material demands of digital infrastructure.
Four gaps stand out.
There is a disclosure gap.
Companies are not required to publish data on energy, water, or land use. Without this information, regulators cannot understand the real scale of resource consumption.
This directly creates an assessment gap.
Environmental reviews such as the Environmental Impact Assessment (EIA) do not capture the intensity or unique risks of AI facilities.
For instance, data centres and compute farms do not appear in the EIA Rules of 2006 and are often treated as regular construction projects, even though their environmental impacts are very different.
These weak assessments lead to an enforcement gap.
There are no clear standards for renewable energy sourcing, energy efficiency, or water use, and no consistent way to enforce them.
Finally, limited public participation makes oversight even harder.
Communities often learn about new projects only after construction begins, leaving little space for consultation.
AI infrastructure has specific environmental pressures that general industrial rules cannot address.
The growth of this industry outpaces the development of targeted regulation. If these gaps remain, India risks locking in resource-intensive practices that will be costly and difficult to reverse later.
They can also deepen the very vulnerabilities the country hopes to overcome.
(Escandita Tewari is a Research Associate in the Adaptation & Resilience vertical at the Sustainable Futures Collaborative.This is an opinion piece. Views expressed are the author's own. The Quint neither endorses nor is responsible for the same.)