The AI revolution is leaving a very real carbon footprint

Developing artificial intelligence (AI) models consumes a lot of energy, which raises greenhouse gas emissions and generates a carbon footprint

Dave Murray

AI is creating more emissions, but it is also being used to find “smart” solutions to fight climate change and offset the cutting-edge tech’s growing contribution to it.

The AI revolution is leaving a very real carbon footprint

Marcelle Nasr

last update on 28 Dec 2023

The rapid ascent of artificial intelligence does not come without costs, including those it levies on the environment, as its technology rapidly advances and consumption of it grows at pace.

Concern about the expanding carbon footprint of the industry is on the rise amid a broader lack of awareness of its contribution to climate change. The increased use of AI has significant implications for the efforts made by technology firms to reduce their emissions, which now look set to rise in line with the increase in data centres needed to power this latest tech frontier.

Some estimates even suggest that by 2025, AI could consume up to 20% of the world's electricity, contributing approximately about 5.5% of global carbon emissions.

As tech firms develop more AI applications for everyday use, the impact on the environment has been largely ignored. But the issue is critical. The internet alone accounts for 1.6 bn tonnes of greenhouse gas emissions annually. The AI race has not included much thought of how it will increase that, in line with the extra energy it uses.

There are currently no programmes to mitigate the emissions from the electronic tools we use, but there is growing interest in renewable energy as a means to reduce emissions. Connecting renewable energy sources to the large servers that power artificial intelligence is seen by some as insufficient, particularly in the field of digital energy.

The Intergovernmental Panel on Climate Change has warned about the devastating consequences of rising greenhouse gas emissions, emphasising the need to simultaneously expand decarbonisation alongside implementing measures to reduce emissions.

Some estimates even suggest that by 2025, AI could consume up to 20% of the world's electricity, contributing approximately about 5.5% of global carbon emissions.

Big footprint

Reducing the overall carbon footprint – defined as the total emissions of greenhouse gasses and a measure of the impact of human activity on climate change – is a crucial element in the global efforts to combat climate change and prevent its devastating repercussions on the world and its international economies.

When individuals, businesses and industries reduce their own carbon footprints, they contribute to limiting harmful emissions and, consequently, mitigating climate change. Reducing the carbon footprint involves investing in renewable energy sources like solar and wind energy.

It also promotes innovation and development in the fields of sustainable energy and environmental technology. Various metrics measure the carbon footprint at different levels, from individuals to entire nation-states. These metrics vary according to the purposes and goals they serve and depend on a range of factors, from energy consumption and transportation to industrial production, agriculture, and environmental management.

Big numbers

Tracking greenhouse gas emissions is one of the main ways to estimate each country's carbon footprint. International organisations, such as the United Nations and the International Energy Agency (IEA), regularly report on country-level carbon emissions and their developments.

The United States alone accounts for 28% of these emissions, followed by the European Union (23%), Japan (4%), the countries of the Organisation for Economic Co-operation and Development (OECD) (5%), Russia (11%), China (9%), India (3%,), and the rest of the world (18%). Developed countries account for 83% of emissions while developing countries are responsible for only 16%

A recent joint study by Google and the University of California, Berkeley, revealed that the training of the generative artificial intelligence model GPT-3 resulted in carbon emissions equivalent to 552 tonnes, roughly the emissions of a car driving two million kilometres.

Its successor, GPT-4, was trained on a scale 570 larger than the previous version — a number likely to increase with the advancement and proliferation of artificial intelligence capabilities.

Major cloud computing companies such as Amazon Web Services, Microsoft, and Google assert their commitment to the highest possible energy efficiency. Amazon Web Services has pledged to achieve carbon neutrality by 2040, while Microsoft has committed to being a "zero waste, carbon-negative company" by 2030.

Some climate experts also view artificial intelligence as a largely positive development, praising AI as a tool that can improve our understanding of our environmental impact and the effects of climate change.

AI could also reduce carbon emissions

Some climate experts also view artificial intelligence as a largely positive development. The United Nations Environment Programme has praised AI as a tool that can improve our understanding of our environmental impact and the effects of climate change.

Dr. Mohamed Abdul Aal, an expert in economics and artificial intelligence technology, told Al Majalla that it is difficult to identify precise figures or specific results related to artificial intelligence's impact on reducing the industrial sector's carbon footprint in general because the results may vary depending on the industry, company, and technologies used.

Yet, some studies and experiments indicate the potential for using intelligent energy data analysis to improve energy efficiency in the industry by up to 20%.

Many countries have used artificial intelligence in industries to reduce carbon footprints. In China, artificial intelligence has been adopted in numerous major industries, such as steel manufacturing and heavy industries. AI technologies have been used to enhance production processes and improve energy efficiency.

In Germany, artificial intelligence has been applied in the car industry and among small to medium-sized enterprises. It has been used to improve production processes, provide predictive maintenance for machinery and equipment, as well as enhance energy consumption efficiency and reduce emissions.

In the US, numerous companies rely on artificial intelligence technologies in various industries to enhance efficiency, reduce carbon footprint, generate energy from renewable sources, and improve the efficiency of electrical grids.

Nonetheless, Abdul Aal emphasised that artificial intelligence does leave a carbon footprint in certain cases: "In the context of artificial intelligence, the carbon footprint refers to the digital impact left by the AI system."

"It includes all digital activities produced by intelligent applications and systems, such as data usage, behaviour analysis, content delivery, and interaction with users," he explained, adding that: "secure encryption technologies can be applied to reduce the carbon footprint of AI applications. Much research and development is going on to improve privacy and security in this new field."

Professor Alaa El Nahry, Deputy Director of the Regional Center for Space Science and Technology Education for Western Asia (RCSSTE-WA) at the UN, referred in an interview with Al Majalla to statistics compiled in 2020 that ranked China as the world's largest source of carbon dioxide emissions. In that year, China emitted 10,668 million metric tons, followed by the US, with a total of 4,713 million metric tons.

India, Russia, and Japan were next in terms of the size of their emissions. The global annual average carbon footprint per person in 2020 was about 6.3 metric tonnes.

Climate-related forecasting and natural resource management

AI's ability to simulate human intelligence using techniques such as deep learning and its ability to apply the techniques at speed to interpret and analyse vast amounts of data with high levels of accuracy makes it ideal for use in understanding climate change.

It can improve the standard of climate-related forecasting. It can also be used in designing mathematical models for climate variations, ensuring the precision of climate forecasts for natural disasters like floods, thunderstorms, heavy rainfall, landslides, heat waves and cold snaps, hurricanes, droughts, and dust storms or snowstorms.

AI can also design models that quickly adapt to climate changes and provide practical strategies to mitigate their impacts.

AI can also be used in natural resource management through non-traditional methods, including precision farming, to improve agricultural production with lower methane emissions.

It can also develop accurate predictive models for crop needs, how to deal with rainy and dry periods, improve energy distribution and storage systems, and enhance their efficiency. This, in turn, reduces dependence on fossil fuels and greenhouse gas emissions associated with global warming.

And so the immense potential of artificial intelligence in combating climate change cannot be denied. However, it is important to address related challenges and concerns to ensure its effective and safe use.

AI can also be used in designing mathematical models for climate variations, ensuring the precision of climate forecasts for all kinds of natural disasters.

A question of balance

As artificial intelligence advances, the technology sector and policymakers face a pressing question: Can a balance be achieved between AI's transformative capabilities and its substantial carbon footprint?

El-Nahry points out that all countries worldwide must collaborate to keep the rise in global temperature capped at 1.5 degrees Celsius compared with pre-industrial levels.

That will mean avoiding fossil fuels – coal, oil, and gas – and making greater use of new and renewable energy sources, such as solar energy, wind energy, green hydrogen and green ammonia.

El-Nahry noted that Egypt is following this approach, generating electricity from solar power at the Benban station in Aswan (1,465 megawatts, expandable to 2,000 megawatts), in addition to wind energy from the Zaafarana and Gabal El-Zeit wind farms in the Red Sea (1,625 megawatts).

To this end, Egypt is constructing factories to produce green hydrogen and ammonia in the Suez Canal Economic Zone (Ain Sokhna) through Arab and international partnerships.

Yet the AI consultant and digital transformation engineer, Romani Rizkallah, confirmed to Al Majalla that every AI contributes to carbon emissions if traditional energy is used to power it.

"This results in emissions, and after a period, these machines are disposed of and replaced by new ones, causing additional environmental emissions. These devices require cooling and recycling, and AI applications consume a lot of energy," explained Rizkallah.

Egypt is considered one of the leading Arab countries that has devoted considerable attention to action plans to combat climate change. The launch of the National Strategy for Climate Change 2050 resulted from hosting the Sharm el-Sheikh Climate Summit and chairing last year's COP27 conference on the issue.

Egypt has been actively working to implement practical measures to achieve adaptation in affected sectors and mitigate emissions from contributing sectors.

A global problem

According to the International Energy Agency, the primary sources of carbon emissions are electricity generation, industry, and transportation.

The electricity and transportation sectors account for 40% and 21% of carbon emissions, respectively, while the information and communication technology (ICT) sector consumes 4%. Other estimates show its consumption higher, at around 10% of the world's electricity.

As data centres form the backbone of our digital world, their increasing proliferation poses a significant environmental challenge amid concerns about the carbon footprint of the technology sector.

Will the solution come from artificial intelligence, or will things get more complicated in the coming period?

+ / -