Deep Seek, a Chinese AI pioneer, overturned the year long high expectations of the US power market within one day. Previously, the industry widely predicted that artificial intelligence technology would lead to a surge in electricity demand, but this Chinese company's technological leap is rewriting the rules.
Energy efficiency innovation overturns traditional predictions
The AI model developed by DeepSeek has achieved an astonishing leap in energy efficiency, with computing power hundreds of times higher than similar products. This indicates that while achieving the same computational efficiency, energy consumption will significantly decrease, which is expected to significantly reduce the impact of the artificial intelligence industry on the climate. The team of Julien Dumoulin Smith, an analyst at Jefferies Group, pointed out that this breakthrough directly challenges the traditional prediction in the United States that AI will contribute about 75% of the growth in electricity demand by 2035. Stephen Jarvis, a professor of environmental economics at the London School of Economics and Political Science, said frankly that the "huge demand forecast" advocated by power companies earlier was more like creating momentum for grid infrastructure investment, "which can't help but remind people of the reappearance of the tulip foam". Eric Gimon, a senior researcher at the Energy Innovation Think Tank, suggests that energy efficiency improvements have always emerged during the historical data center construction boom, and this time may not be an exception.
The clean energy competition is in full swing
Despite breakthroughs in energy efficiency, the demand for stable power sources in the AI industry remains strong. Microsoft and Constellation Energy have restarted their cooperation on the Three Mile Island nuclear power plant, and Meta has announced measures such as seeking 4 gigawatts of nuclear power supply, reflecting the urgent need of technology giants for all-weather clean energy. Bloomberg think tank analyst Nikki Hsu pointed out that even if the growth rate of AI electricity consumption slows down, residential electrification and industrial decarbonization will continue to drive overall electricity demand growth. Behind this energy anxiety lies the increasingly deviated climate commitments of tech giants. Microsoft's carbon emissions in 2023 have surged by 30% compared to 2020, while Google's emissions have increased by 48% compared to 2019, with AI computing power expansion being the main reason. Sweden expects its electricity consumption to double this decade, while the UK predicts a 500% increase in demand during the same period. Although these forecasts may be lowered, the pressure of energy transition remains enormous.
Technological innovation and energy consumption dilemma
The lightweight modeling technology adopted by DeepSeek provides a new perspective for reducing AI carbon emissions. However, Bloomberg analyst Nikki Hsu reminds that key technical details such as chip energy efficiency still need further validation. What is even more alarming is the shadow of the Jevons paradox - efficiency improvement may stimulate wider applications, but instead lead to an overall increase in energy consumption. This competition of innovation and consumption has escalated into an international competition. The $500 billion "Stargate" project recently promoted by Trump stands in stark contrast to DeepSeek's technological breakthrough. Benton Arnett, Senior Director of Policy Markets at the Nuclear Energy Research Institute, emphasized that "if the United States wants to maintain its leading position in AI, it must ensure the energy supply of data centers." Most nuclear energy projects will not adjust their plans due to short-term fluctuations. This cross Pacific AI energy game is reshaping the future landscape of the global electricity market. With technological innovation constantly challenging physical limits, humanity is seeking a delicate balance between computing power and sustainable development.