Artificial Intelligence (AI) technology is advancing rapidly and garnering significant attention. There are high expectations that AI will contribute to addressing global warming through improved weather prediction, climate modeling, and exploration of carbon reduction technologies on a global scale. But can AI really solve the climate crisis?

Is AI Actually Exacerbating the Climate Crisis?

Global environmental organizations like Greenpeace and Friends of the Earth (FOE) argue that AI is not helping to solve the climate crisis, but rather making it worse.

Developing and using AI promotes massive energy consumption. This is particularly severe for generative AI like ChatGPT. Training and using generative AI for images and videos requires enormous data transfers to cloud servers. So, why is AI receiving as much criticism as it does praise for its potential?

Problem 1: Power Consumption

When we ask ChatGPT a question, it consumes 2.9Wh of electricity. This is ten times higher than the average power used for a Google search (0.3Wh). Image generation AI requires enough energy to charge a smartphone for each image it creates.

The power consumption for AI training is staggering. ChatGPT's training alone used 1287MWh of electricity, equivalent to the annual usage of 120 households in the United States. This translates to 520 tons of CO2 emissions, similar to what 110 US cars emit in a year.

According to the International Energy Agency (IEA), in 2022, when generative AI began to be developed and commercialized in earnest, the world's data centers consumed 460 terawatt-hours (TWh) of electricity annually. This is more than double the 200TWh consumed in 2015. The IEA predicts that by 2026, power consumption will exceed 1000TWh, more than double that of 2022. This raises concerns about an 80% increase in greenhouse gas emissions, as more fossil fuels will be needed to meet the enormous power demands of data centers.

Big tech companies are focusing on building data centers to prepare for future power shortages. For example, Amazon acquired a data center powered by a nearby nuclear power plant for $650 million in March, and NVIDIA CEO Jensen Huang announced on March 19 that he would invest $1 trillion (about 1,339 trillion won) in building data center infrastructure. In other words, both the international community and big tech are anticipating massive power consumption in the future.

Problem 2: Water Issues

AI operates based on cloud computers in data centers. A data center consists of servers with thousands of semiconductors. Naturally, it uses enormous amounts of power and generates just as much heat. Think of the server room that appears in the movie 'Mission Impossible'.

Can you imagine how hot it would be inside if each of those small boxes was equivalent to a desktop computer's main unit? Most server rooms have cooling facilities to manage this heat, namely cooling water.

Cooling water has several requirements, notably that seawater cannot be used to prevent corrosion and bacterial growth. Cooling water must be clean freshwater.

So, how much water is AI consuming?

For ChatGPT, one conversation with 25-50 questions and answers consumes 500ml of water. It's estimated that training GPT-3 alone required 700,000 liters of water, equivalent to producing 370 BMW cars.

Moreover, from 2021 to 2022, Microsoft's water consumption was 6.4 billion liters, enough to fill over 2,500 Olympic-sized swimming pools. Big tech companies are consuming water at a rapid rate as they invest in AI.

Building a single data center requires an enormous amount of cooling water. Is it okay to use so much freshwater that's scarce even for drinking? Of course not, especially in water-stressed regions. In 2023, Google tried to establish a data center in Uruguay but had to scale back its plans due to residents' opposition. At the time, Uruguay was experiencing its worst drought due to abnormal weather, making it difficult to even secure drinking water. When news broke about the establishment of a data center that would use the daily water consumption of 5,000 households as daily cooling water, citizens staged large-scale protests, and Google eventually retreated, reducing the scale of the data center. This is not just happening in Uruguay but all over the world.

A Sobering Reality

Does AI still seem like a panacea for humanity's climate crisis?

While AI provides many benefits and conveniences to humans, it's also consuming enormous amounts of energy and resources. Electricity and water are particularly crucial resources for humanity, making this consumption critical. The bad news is that as AI develops and penetrates deeper into society, the amount of resources consumed is expected to increase exponentially.

Isn't it chilling that the development of technology, which was supposed to save the world, is instead choking humanity?

However, not everyone is sitting idly by in the face of this situation. The world must continue to move in a better direction. In the next part, we'll explore how these crises are being addressed.

Stay tuned for the next installment!