Artificial intelligence technologies have long been a topic of discussion regarding their environmental impacts due to their increasing energy consumption. Particularly, large language models can place a significant burden on global energy consumption as they require vast computing resources. DeepSeek, based in China, claims to have developed an artificial intelligence model that consumes less energy. However, the Jevons Paradox, proposed in the 19th century, suggests that increases in efficiency do not always reduce energy consumption; rather, they may increase demand for energy in the long run. This article explores DeepSeek's claim and the Jevons Paradox.
How Much Energy Do Artificial Intelligence Models Consume?
Today, large language models respond to users' queries similarly to search engines. However, unlike traditional search engines, artificial intelligence models generate their answers from scratch. This process is considered quite energy-consuming as it requires a large amount of computational power.
According to research, the artificial intelligence sector is expected to consume between 85 and 134 terawatt-hours (TWh) of electricity by 2027. This amount corresponds to the total energy consumption of the Netherlands for a year. Experts predict that by 2030, more than 20% of the electricity generated in the U.S. will be directed to artificial intelligence data centers.
Although technology giants claim to be making significant investments in renewable energy sources, they are making more stable energy solutions, such as nuclear power, essential for providing uninterrupted power to artificial intelligence systems. For instance, Microsoft plans to reactivate the Three Mile Island plant, known for being one of the largest nuclear accidents in U.S. history. Meanwhile, although Google aims to be carbon neutral by 2030, carbon emissions in the artificial intelligence sector have reportedly increased by 48% in recent years.
What Is DeepSeek's Claim?
DeepSeek claims to have found a solution to this problem. The R1 model developed by the company was trained at a significantly lower cost compared to its major competitors. While Meta spent over $60 million on its Llama model, DeepSeek reportedly achieved similar performance with just a budget of $6 million. DeepSeek states that its model uses a machine learning architecture called Mixture of Experts, thus allowing for more efficient operation.
These developments have also impacted the global economy. In the U.S., stocks of chip manufacturers and energy companies experienced significant declines. Nvidia, a producer of artificial intelligence processors, faced the largest drop in Wall Street history, losing $589 billion in a single day. Interestingly, improving the energy efficiency of artificial intelligence may lead to an increase in the overall energy consumption of the sector rather than a decrease. This is because the Jevons Paradox argues that more efficient energy use can lead to increased demand for energy.
What Is the Jevons Paradox?
The Jevons Paradox was first proposed by William Stanley Jevons in 1865. Jevons argued that the increased efficiency of coal usage did not reduce consumption, but rather increased it. According to Jevons, more efficient systems led to lower production costs, resulting in increased investments and thereby raising total consumption. The same situation applies today. Technological advancements may provide energy savings, but they do not decrease overall energy demand.
Conclusion
The increasing energy consumption of the artificial intelligence industry continues to be a significant issue regarding environmental impacts. Although more efficient artificial intelligence models developed by companies like DeepSeek may appear to reduce energy consumption in the short term, the Jevons Paradox indicates that this increase in efficiency may lead to an overall increase in total consumption in the long run. Lower costs and reduced energy consumption may lead to a surge in demand as artificial intelligence becomes more widespread and accessible.
At this point, it is crucial for companies not only to increase energy efficiency but also to integrate sustainable and renewable energy sources. However, given the pace of technological advancements and the rise in demand, it does not seem likely that artificial intelligence will be able to permanently reduce its energy usage. As a result, the demand for energy will continue to rise, necessitating the development of more comprehensive solutions to mitigate environmental impacts.
If you are looking for a system that will consume as few resources as possible while automating efficiently, let’s examine your system together. With PlusClouds’ AI-based automation solutions, such as Eaglet and Leo, you can optimize your business processes while saving both time and labor. These intelligent systems can increase operational efficiency by up to 5 times, minimizing unnecessary resource consumption.
Eaglet identifies potential customers within seconds using advanced data scanning technology, analyzing contact information and scheduling your meetings. It can also be used for SWOT analysis and determining competition. On the other hand, Leo fully automates the management of your data center, offering end-to-end control from a single panel. This way, you not only reduce costs but also minimize your environmental impact. To have a more efficient, sustainable, and smart infrastructure, contact us and let’s determine the most suitable automation solutions for you!
