China-based artificial intelligence (AI) startup DeepSeek has sparked a debate over energy consumption in artificial intelligence, claiming that its advanced AI model is far more energy-efficient than those of its competitors.
The announcement challenges existing assumptions about AI's increasing energy demand, prompting investors to reassess their positions in energy markets.
Major tech firms and data centers had expected that AI's growing computational demands would strain global energy supplies. Large-scale AI training models have so far been seen as consuming as much electricity as entire cities.
However, DeepSeek's claim that its models operate with significantly lower energy consumption has cast doubt on these projections. The company's announcement has also shifted market expectations regarding AI investments with high energy demands.
- Energy stock shares drop
Following DeepSeek's announcement, investors began to question AI's projected energy-driven expansion, triggering a sharp sell-off in energy stocks.
Texas Pacific Land, a key landholder for data centers, saw its shares fall 7.7%, as its competitor, LandBridge, dropped 17%.
EQT and Antero Resources, both natural gas producers, fell over 9%, while pipeline operators Kinder Morgan and Williams Cos. declined 9.3% and 8.4%, respectively.
Uranium miners supplying nuclear plants also suffered losses, with shares dropping 11%.
Independent power producers Constellation Energy and Vistra, among last year's top-performing energy firms, plunged 21% and 28%, respectively, after DeepSeek's announcement.
Concerns over AI's energy footprint remain. Microsoft's carbon emissions rose 30% since 2020, while Google's emissions increased 48% since 2019, largely due to AI-driven power consumption.
AI-related electricity demand is expected to double in Sweden and increase 500% in the UK over the next decade.
DeepSeek's more energy-efficient AI model is now being viewed as a potential alternative to mitigate AI-driven emissions.
- DeepSeek's AI model changes demand expectations
Prof. Dr. Serhat Yuksel, head of the Economics and Finance Department at Istanbul Medipol University told Anadolu that the new development can be seen as a radical shift.
Referring to that fact that lower energy consumption reduces costs while maintaining high performance, Yuksel said, 'This fundamentally changes expectations regarding energy demand, particularly affecting projections for data center energy consumption.'
He noted that prior to DeepSeek's announcement, there was a strong expectation in the US that AI advancements would drive up energy demand, leading energy producers to gain substantial value.
However, the announcement has challenged these assumptions, Yuksel said, adding, 'This extraordinary development has led to significant declines in energy stocks.'
Yuksel pointed out that the market impact extended beyond energy producers, affecting natural gas suppliers and energy infrastructure firms.
He highlighted that DeepSeek's AI model offers a significant difference in energy consumption compared to existing models, incorporating optimized algorithms and more efficient data processing techniques to reduce power usage.
'These innovations enable AI models to perform the same tasks with lower computational power, significantly reducing electricity consumption,' Yuksel explained.
- DeepSeek's AI model could boost renewable adoption
The development of low-energy AI models like DeepSeek could significantly reduce the need for expanding data centers and energy infrastructure, according to Prof. Dr. Hasan Dincer, head of the Banking and Insurance Department at Istanbul Medipol University.
Dincer stated that reducing computational power requirements would result in lower hardware usage and energy consumption.
'Some regions are already struggling with supply-demand imbalances in electricity,' he said.
By consuming less energy, DeepSeek's AI model could ease pressure on power grids and reduce the need for further energy infrastructure expansion, he explained.
Dincer also noted that AI's high energy consumption has hindered some data centers from transitioning to fully renewable energy sources.
'With DeepSeek's lower energy requirements, it may become feasible for data centers to operate entirely on solar and wind power,' Dincer added.
By Gokce Topbas
Anadolu Agency
energy@aa.com.tr