By Dr Niladri Choudhuri
The rise of artificial intelligence (AI) has revolutionized industries, from healthcare and finance to autonomous vehicles and entertainment.
In a two-part series, Dr. Choudhuri writes about how Al-driven data centers are guzzling water and power and offers some green alternates.
However, this technological leap comes at a significant cost to the environment, driven largely by the escalating resource demands of data centers that power AI computations. These facilities, integral to modern digital infrastructure, are notorious for their immense energy and water consumption, alongside their considerable carbon footprint.
Power and water consumption
Data centers worldwide consume approximately 200 terawatt-hours (TWh) of electricity annually, accounting for about 1% of global electricity use. This figure is steadily climbing due to the exponential growth of AI workloads, which are computationally intensive and require advanced hardware like GPUs and TPUs. Training a single AI model can consume as much power as 100 average US homes use in a year.
Water usage is another critical issue. Data centers rely heavily on water for cooling purposes. Estimates suggest that a large-scale data center can consume 300,000 to 1 million gallons of water per day, equivalent to the daily water needs of a small city. This dependency poses a severe risk in regions prone to water scarcity, exacerbating local environmental stress.
Carbon emissions
Data centers contribute significantly to greenhouse gas emissions. The annual carbon footprint of global data centers is estimated to be around 200 million metric tons of CO2, equivalent to the emissions from 43 million cars on the road. With the rapid adoption of AI, this figure is expected to grow unless sustainable solutions are implemented.
The environmental implications are clear: the traditional approach to building and operating data centers is unsustainable. The urgent need for greener alternatives is driving innovation in data center design and energy solutions.
Micro data centers: a green alternative
As the demand for data processing grows, microdata centers (MDCs) are emerging as a promising solution to mitigate the environmental impact of traditional large-scale data centers. These compact, self-contained units are designed to deliver localized computing power with significantly reduced resource requirements.
What are MDCs?
These are smaller-scale versions of traditional data centers, typically housed in a single enclosure. They include all essential components: computing, storage, networking, and cooling systems. Designed for deployment closer to the end user, MDCs support edge computing, enabling faster data processing and reduced latency. They have an energy need of 25 KW to 300 KW. Many of these do not require water for cooling at all.
Benefits of MDCs:
- Energy efficiency: MDCs are designed to operate efficiently in constrained environments. They often incorporate advanced cooling technologies, such as liquid cooling or closed-loop air systems, that significantly reduce energy consumption. Some have a PUE of 1.2 or less.
- Reduced water usage: Unlike traditional data centers, which rely on water-intensive cooling systems, many MDCs use air or liquid cooling solutions that require minimal to no water, making them a greener choice. Some work on domestic air conditioner systems, and the centre can run at 22 degrees Celsius and hence does not need much cooling.
- Localized deployment: By processing data closer to the source, MDCs reduce the need for long-distance data transmission, lowering the associated energy costs and carbon footprint.
- AI Compatibility: Despite their smaller size, MDCs can be equipped with high-performance computing hardware to handle AI workloads efficiently. Modular designs allow for scalability based on specific computational needs. We really do not need huge computing resources to run AI in every use case. Most of the AI needs can be handled with much smaller compute needs with proper designing.
- Green design: Many MDCs incorporate renewable energy sources like solar panels or wind turbines and biomass energy, further minimizing their environmental impact. Their compact size also enables integration into urban or remote environments, where traditional data centers may not be feasible.
A sustainable step forward:
Microdata centers represent a shift toward decentralization, aligning with the principles of sustainable IT. By reducing the reliance on hyperscale DCs, MDCs contribute to a more balanced and environmentally friendly digital infrastructure.
Dr. Niladri Choudhuri is the founder and CEO of Xellentro Consulting Services LLP and President of the Green Computing Foundation
Next Week: Biomass Energy: A Green Power Source for Micro Data centers
