Did you know MIT reported that the cloud has a larger carbon footprint than the entire airline industry?
As big data, artificial intelligence (AI), and data storage in the cloud become increasingly prominent in today’s digital age, concerns about the environmental impact of data have risen.
The rapid growth of data centers, the energy consumption required for computation, and the carbon emissions associated with data storage have all contributed to the staggering ecological impacts of computation and the cloud.
The Staggering Environmental Impact of Data and the Cloud
Storing 1 terabyte of data in the cloud has a carbon footprint of 2 tonnes annually.
Data centers, the backbone of the cloud, require extensive material infrastructure, including cables, HVAC, lights, emergency power banks, and computer servers.
At its worst, only 6-12% of the energy consumed is for active computational processes, whereas the remainder is for maintaining the temperature and multiple chains of fail-safes to prevent any downtimes.
Cooling, which accounts for over 40 percent of electricity usage in data centers, leads to significant energy consumption, and the cloud’s carbon footprint surpasses that of the airline industry.
Though the shift from HVAC cooling to water cooling has been an attempt to reduce the carbon footprint, it still comes with its own set of problems. Data centers also consume vast amounts of water for these cooling processes, putting strain on water-stressed communities, as seen in regions like Arizona and Utah.
As a result of these data centers, the residents of Buffdale, Utah have been suffering from water shortages and power outages, thanks to a nearby U.S. National Security Agency (NSA) data center that consumes seven million gallons of water daily to operate.
Furthermore, data centers need to stay clean so pollutants like dust don’t disrupt their technological functions. To keep them to these standards, toxic cleaners are used and old batteries are disposed of, which adds to the toll that data has on the environment.
The manufacture of information and communication technology (ICT) devices, including smartphones, also contributes to environmental issues through the use of fossil fuels, chemicals, and water, contributing to e-waste.
While efforts are being made to reduce e-waste through equipment recycling programs, the current reality is that progress remains slow. The ecological impacts of the cloud are not predetermined, and they depend on human practices, choices, and efforts to make data storage and computation more sustainable.
The Carbon Footprint of Data Storage
Data storage, often referred to as “the cloud,” offers convenience and easy access to files across multiple devices. The energy used by data storage centers and associated devices contributes to 2% of global carbon emissions.
Despite their impact, the reality is that data and AI aren’t going anywhere, so we must mitigate the impacts as effectively as possible.
Sustainable Alternatives and Strategies
Addressing the environmental impact of data storage and computation requires innovative strategies and sustainable alternatives.
The essential pillars of consideration when making AI more sustainable is to examine how and where data is stored, and increase transparency and measurement to allow for accurate insight into carbon emissions to inform sustainable innovation.
Carbon Accounting and Footprint Estimation:
One suggestion is to assess how the environmental impact is measured in improving carbon accounting.
Improving carbon accounting by delivering faster, more accurate data on carbon footprints and sustainability impacts helps companies visualize and understand their environmental impact, enabling them to spot opportunities for improvement.
The Machine Learning Emissions Calculator can help estimate the carbon footprints of AI models, leading to more sustainable decision-making.
Energy Efficiency in Data Centers:
Data centers can implement techniques to eliminate storage waste, including data compression, deduplication, thin provisioning, and efficient use of virtual machines.
Reducing energy consumption through more efficient cooling systems and cleaner energy sources can significantly decrease data centers’ carbon emissions.
Public cloud data centers that shift towards renewable energy sources can see a benefit of carbon intensity reduction by up to 84%. (Source: The Economist)
According to a Lawrence Berkeley National Laboratory report, if Cloud computing shifted to hyper-scale facilities, energy usage might drop as much as 25% as well.
Sustainable Location and Water Consumption:
Relocating data centers to cooler regions such as Nordic countries like Sweden, or Iceland with access to renewable energy sources can help reduce their environmental impact.
Using the naturally cool air in these landscapes (a technique known as “free cooling”) may minimize carbon footprints, however, network signal latency issues in remote areas cause issues when in the relocation of data centers.
Additionally, data center companies should prioritize water efficiency and consider “water-positive” initiatives to replenish more water than they consume, supporting water-stressed communities.
Large-scale companies like Google and Cyrus One can utilize a “closed-loop” water cooling system that recycles some of the wastewater used in evaporative cooling, committing to go “water-positive” by 2030 and replenish 120% of the water they consume.
Sustainable Development Practices:
According to MIT Review, between 70% and 90% of data organizations collect is considered “dark data” – meaning data that is not turned into insights and business opportunities – costing companies unnecessary energy costs to transmit and store this valueless data.
The nature of data emphasizes the importance of cleaning and preparing your data.
This not only ensures that the data used is relevant and is in the correct format, but it focuses on quality over quantity concept, ultimately reducing silos. Reducing any storage waste using “lean principles” (as seen in manufacturing) means a win-win for your business.
This is because utilizing smaller data sets can both reduce costs and offer granular business decision-making insights while simultaneously minimizing environmental footprints required for maintenance and management (Source: The Economist).
Sustainable data processing involves labelling data at the beginning, optimizing data transfer, and purging and archiving data using energy-efficient storage options.
Creating machine-learning models with limited representative data ensures more sustainable AI practices.
Techniques such as encoding, quantization, and pruning, can aid in making even the most energy-intensive AI models more sustainable.
Emphasizing Transparency and Measurement:
AI researchers should publish results for new models, including measurements of energy emissions alongside performance metrics. This allows for better transparency about energy consumption and carbon emissions.
Keeping AI companies accountable fosters a community that focuses on sustainability, and allows for better evaluation and comparison of AI models.
As AI continues to gain prominence in our world, it is important that we address these influences before greater repercussions occur.
Following Best Practices:
Adopting best practices like Google’s “4M” approach, which focuses on selecting efficient machine learning model architectures, using optimized processors and systems for training, computing in the cloud, and choosing locations with clean energy, can significantly reduce energy and carbon emissions.
The Growth of Data isn’t Stopping – So We Need To Address it Now.
As data and AI continue to grow, the environmental impact of data storage, computation, and AI is a pressing concern that requires immediate attention.
The exponential growth in data and AI deployment has contributed to a staggering ecological impact, with data centers and the cloud consuming vast amounts of energy and producing significant carbon emissions.
However, through sustainable practices and strategic solutions, such as improving carbon accounting, optimizing energy consumption in data centers, and prioritizing renewable energy sources, it is possible to mitigate the environmental impact of data storage and computation.
By taking bold and thoughtful initiatives to make data and AI more sustainable, organizations can achieve increased resilience, lower costs, and higher efficiency, and contribute to a brighter future for all stakeholders. The responsibility to transform the way we handle data lies with us, and by adopting sustainable practices, we can ensure a more sustainable and environmentally friendly digital future.
Ready to unlock your organization’s full potential? Contact us today and transform your organization’s data challenges into opportunities.
