AI, Sustainability, & the Crossroads of Progress

As artificial intelligence (AI) becomes embedded in daily activities, it comes with measurable environmental costs. 

AI banks on a growing physical infrastructure of data centers, advanced chips, cooling systems, electricity grids, water resources, land, and critical mineral supply chains. 

The benefits of AI often flow across borders and sectors, while the environmental burdens need to be addressed. 

Innovations such as high-density GPUs (Graphics Processing Units) and specialized AI chips are reshaping infrastructure requirements for AI. GPUs are now central to modern computing, especially in AI, data analytics, and high-performance computing. 

India’s initiative to set up 18,000 advanced GPU-based facilities as part of the IndiaAI Mission, in early Q4 of Fiscal 2025, is expected to greatly increase power consumption within the data center industry. The data center infrastructure is also demanding AI-powered cooling systems. This will require thermal loads needed to mitigate energy consumption.

AI workloads significantly increase power demand and to manage this, data centers are adopting high-density cooling systems and resilient power architectures. 

“India’s power landscape is supportive with total generation capacity at 452.69 GW and renewables accounting for more than 46 per cent of installed capacity. As of mid-2025, India has effectively achieved zero power deficit, ensuring reliable supply. Our Data centres use substantial energy, so improving efficiency and reducing environmental impact is important. Our largest facility uses close to 60 per cent renewables in its mix. We are also increasingly adopting renewable energy sources like solar, wind, and bioenergy for sustainability,” said Sharad Agarwal, CEO, Sify data center.

One of the most significant shifts will be the adoption of advanced cooling technologies such as direct-to-chip liquid cooling and immersion cooling, which are essential to manage rising power densities in GPU-intensive environments.

Power architecture will also evolve, with higher-capacity electrical systems, modular power distribution, and integration of renewable energy sources to meet sustainability goals. Intelligent energy management systems and improved power usage effectiveness will become key differentiators.

Building on this foundation, AI-enabled automation will transform facility operations through predictive maintenance and real-time workload balancing. To support this new level of agility, modular and prefabricated data center builds will enable faster deployment and flexible scaling to meet dynamic demand.

Economic growth always extracts a price, whether in power, water, or other natural resources.

“There will always be a need for balance between growth and the concerns for climate. Any economic growth in any sector will always come at the expense of resources, whether they are consumed by machines or buildings or humans, there will be a cost,” said Sunil Gupta, founder, Yotta Data Services.

Data centers are power guzzlers, and now face amplified scrutiny with the rise of AI and GPU-driven workloads. “However, in India, for at least the next decade, the link between AI expansion and resource depletion will remain manageable. The challenge is real, but not immediate,” added Gupta.

Power is no longer an afterthought, it’s a design constraint. From transistors to firmware, and all the way up to cloud orchestration, every layer must now be engineered with sustainability at its core. What was once an optimization problem has become a baseline requirement. “GPUs, once optional, are now indispensable, and with that shift, sustainability has moved from choice to necessity,” said Ananda Bhattacharjee, head, AI Enterprise, Lenovo Asia Pacific.

Global data centres consumed an estimated 448 terawatt-hours of electricity in 2025.

According to the United Nations University Institute for Water, Environment and Health, discussion has largely focused on the energy required to train massive models. Training GPT-3 was estimated to require 1.3 gigawatt-hours (GWh) of electricity, while estimates suggest GPT-4 consumed between 50 and 70 GWh. ChatGPT alone is estimated to process around 2.5 billion prompts per day, translating to roughly 383 GWh of electricity per year for a single product. 

“Offsetting associated carbon emissions would require 2.6 million tree seedlings grown for 10 years, enough trees to cover a land area the size of Manhattan. The water footprint is equivalent to the minimum annual domestic water needs of roughly 500,000 people in Sub-Saharan Africa, and the land footprint is equal to over 800 football fields,” the report added.

As AI becomes inseparable from modern infrastructure, its environmental footprint can no longer be treated as collateral damage. Building a responsible AI ecosystem demands more than incremental fixes. It requires a framework across transparency, efficiency by design, equity and environmental justice, lifecycle responsibility, global cooperation, and sustainable use. Together, these principles chart a path where innovation and sustainability are not opposing forces but interdependent goals.