The advent of AI is reshaping industries, with the electric power sector poised to be one of the first to benefit significantly.
Recently, electric power stocks have seen a considerable uptick in their market valuesNotably, the State Grid and Southern Power Grid have announced an unprecedented investment of 600 billion yuan in 2024, marking a historic highThis surge in investment is a crucial factor bolstering the stock prices of related companies, alongside the accelerated pace in Chinese manufacturing and the growing demands of AI computing powerThe success of recent AI innovations, specifically DeepSeek, has triggered a significant increase in power needs, instigating substantial changes on the supply and demand frontAs a result, the capital markets maintain a positive outlook on the electric utility systems.
In 2024, China's national electricity system has delivered impressive results, completing the operation of three ultra-high-voltage projects and achieving a cumulative total of 38 ultra-high-voltage constructions, including “22 alternating current and 16 direct current” projectsOver the year, 43,800 kilometers of lines at 110 kV and above have been put into operation, entirely accelerating the quality development of the power gridThe annual investment exceeding 600 billion yuan has been directed toward the establishment of ultra-high-voltage alternating and direct current engineering, as well as the digital and intelligent upgrading of the power grid.
Recently, the State Grid announced plans to increase investment dramatically by 2025, targeting an investment volume exceeding 650 billion yuan to optimize the main power grid, reinforce the distribution network, and effectively serve the high-quality development of renewable energy
Advertisements
Similarly, Southern Power Grid is set to allocate a fixed asset investment of 175 billion yuan, marking a historical peak.
Industry experts believe that investments in the power grid represent crucial methods for stabilizing economic growth and serve as key elements for counter-cyclical adjustmentsThis dual strategy not only meets the surging electricity demands but also creates opportunities for downstream enterprises across various sectorsFurthermore, it provides assurance for inter-provincial and inter-regional electricity transactions, contributing to the establishment of a unified national electricity market while supporting the AI sector; a pivotal element in boosting China's manufacturing and intelligent constructionGiven the significant increase in investments alongside robust market supply and demand, the electric power system could witness unprecedented growth come 2025.
Electricity serves as a litmus test for a nation or region's developmentHistorically, China has borne the label of a “weak power nation.” By 1989, China's electricity generation just caught up to the levels achieved by the United States in 1953, reflecting substantial historical gaps as average per capita electricity consumption was less than one-tenth that of the United States. Over the years, by 2003, China's electricity generation reached about half that of the USA.
In 2011, China achieved a major milestone, becoming the world's largest electricity producer, generating a staggering 47,306 billion kWh, and overtaking the USA in installed power capacity.
In 2014, the USA generated 40,935 billion kWh, with a modest growth of 0.7%. In comparison, China's generation reached 55,495 billion kWh, marking a growth rate of 3.8%. Thus, China's electrical output became 1.36 times that of the USA, while China’s GDP for that year stood at $10.48 trillion compared to the USA's $17.61 trillion, translating to 59.5% of the American economy. However, China’s installed power capacity began to grow at an even faster rate in subsequent years, with the cross-regional grid capacity being 80 times that of the USA between 2014 and 2021.
In 2014, the electricity generation structure in the USA was predominantly natural gas and other gaseous sources, accounting for 42.4%. Coal-fired power contributed 27.8%, while nuclear power made up around 8.9%. Meanwhile, wind and photovoltaic energy were seen to grow significantly, at 5.6% and 0.9% respectively
Advertisements
While exact numbers for China were not detailed, the nation's efficient construction processes and rapid energy production indicated ongoing optimization and modernization.
In 2024, the electricity output in China was recorded at 9.4 billion kWh, while the total generation in the USA stood at 4.3 billion kWh, clearly showing China's electricity generation was about 2.2 times that of the USAThe total installed capacity was also revealing: China boasted 2.6 billion kWh compared to the USA's 1.3 billion kWh, affirming that the USA’s total installed capacity was half that of China's.
In 2024, the contribution of coal-fired power in the USA was still above 15%, exceeding the total of all renewable energy sourcesWind power claimed an 11% share, whereas hydropower and solar energy combined reached merely 10%. These figures lag significantly behind those of China, where the share of renewable energy generation approached 50%, and the utilization rates for wind and solar power soared above 95%. In 2024, nearly 80% of the increase in national power generation derived from renewable energy.
In 2024, the GDP of the USA reached $29.2 trillion, reflecting a growth rate of 2.9%, while China's GDP was $18.94 trillion, at a rate of 5%. China's GDP accounts for 65% of that of the USA; however, adjusting for currency differences suggests that this gap is narrower than it appearsThe rapid acceleration of China's economy and the reduction of its distance from the USA owe much to its robust electricity supply.
While AI technology progresses exponentially, it is often overlooked that the power consumption required for AI computing is substantial
Advertisements
This sector relies heavily on electricityReports state that global data centers consume approximately 1%-2% of the total global electricity demandAs AI computing needs grow, this proportion is expected to rise significantly.
AI practitioner Wang Anran highlights the massive electricity consumption associated with AI power, particularly during the training and execution of large-scale deep learning modelsAs AI models scale (like GPT-3 and GPT-4), their computational resource requirements increase exponentially, leading to a sharp rise in power consumption.
“AI computing is divided into the training phase and the inference phaseTraining large-scale AI models necessitates extensive computational resources, typically utilizing high-performance GPU or TPU clustersThe iterative nature of training necessitates immense computations, leading to exceedingly high power consumptionFor instance, training a model like GPT-3 may take weeks, consuming hundreds of thousands of kilowatt-hours of electricityAlthough the computational load during the inference phase is comparatively lower, the sheer scale of users keeps total power consumption highApplications like ChatGPT handle millions of requests daily and require numerous servers for ongoing operationsMoreover, the computational devices, cooling systems, and related network equipment in data centers consume substantial electricity,” Wang Anran emphasizes that while there are challenges lowering energy efficiency today, doing so will be essential for future advancements.
Currently, GPT-3 possesses 175 billion parameters, and its training process consumed around 1,287 megawatt-hours (MWh) of power—enough to power 121 American households for a yearTraining ChatGPT equates to the energy used by 3,000 Tesla electric cars driving together for 200,000 miles each, while the larger GPT-4 model is expected to consume several times more power during training.
While specific data regarding GPT-4's needs have yet to be disclosed, estimates suggest it could demand thousands of megawatt-hours of power
For ChatGPT, each interaction consumes about 2.9 watt-hoursCurrently, it processes approximately 195 million requests each day, consuming about 564 megawatt-hours daily, equivalent to the consumption of 17,000 American households—a clear indication of its significant energy appetite.
"The deep training of AI requires massive amounts of data input, relying on high-end GPUs or custom AI processors to enhance computation speedsThe growth of data volume directly leads to increased computational demands, meaning that these high-performance computing components consume vast amounts of energyExtensive memory access and computational operations further elevate energy consumption," Wang Anran explains the heavy energy consumption rationale of AI computing. “Lowering energy use can only be achieved by optimizing algorithms to decrease computation needs, developing specialized AI chips for improved energy efficiency, and utilizing renewable energy sources to power data centersCurrently, companies are experimenting with these approaches, but results are not yet significant.”
Electric power propels the growth of AI computing capabilities, and in turn, AI indirectly supports the power industryAI algorithms can ensure precision in supply and demand on the generation sideIntelligent power control systems finely adjust generator operating parameters and optimize control strategies, significantly increasing generation efficiency and system stabilityAt the network level, AI algorithms enable intelligent power dispatch systems to monitor real-time grid status and optimize logistics, ensuring that supply and demand remain balanced while enhancing overall grid operating efficiencyConsumer-side applications lead to cost reduction and operational efficiency, directly reflected in power companies' financial reports.
For instance, State Grid's subsidiary, Guodian Nari, reported a revenue of 32.31 billion yuan in the first three quarters of 2024, marking a growth of 12.97% year-on-year, with net profit attributable to shareholders reaching 4.473 billion yuan, up 7.53%. The revenue increase represents a three-year high for the same period
Between 2021 to 2023, the revenues were 42.41 billion yuan, 46.83 billion yuan, and 51.57 billion yuan, respectively, reflecting growth rates of 10.15%, 10.42%, and 10.13%. Net profits for those years stood at 5.642 billion, 6.446 billion, and 7.184 billion yuan, with consecutive growth rates of 16.30%, 14.24%, and 11.44%. Such impressive financial performance resulted in a 17% increase in Guodian Nari’s stock price for 2024, outperforming the marketHowever, in 2025, Guodian Nari's stock price saw a decline of around 7%.
Noticeably, Guodian Nari's net profit growth has consistently outpaced its revenue growth in the past three years, reflecting an increase in profit marginsFrom 2021 to 2023, Guodian Nari’s gross margins remained relatively stable at 26.88%, 27.04%, and 26.80%; however, net profit margins increased steadily, recorded at 14.25%, 14.74%, and 14.83%. This steady improvement in net profit margin has led to yearly increments significantly larger than that of revenue.
Not every power company's performance has been as stellar; for example, the business of State Grid Yingda has faced volatilityAccording to financial reports, revenue figures from 2021 to 2023 were 9.485 billion yuan, 10.86 billion yuan, and 10.9 billion yuan respectively, with year-on-year growth rates showing a constant slowdown at 17.22%, 14.51%, and 0.32%. A deeper analysis showed net profits were also fluctuating during the same period, standing at 1.224 billion yuan, 1.101 billion yuan, and 1.364 billion yuan, with transformations exhibiting growth percentages of 40.66%, -10.07%, and 23.96%. Over the last six complete financial years, State Grid Yingda has experienced a pattern of alternating upward and downward trends in net profit growth.
In the first three quarters of 2024, revenues at State Grid Yingda amounted to 7.694 billion yuan, reflecting a year-on-year increase of 4.64%, while net profits reached 1.452 billion yuan, up 26.26%. Although the revenue growth surpassed the previous year’s figure of 1.85%, net profit growth was considerably lower than the 39.78% increase noted the prior year
Overall, State Grid Yingda demonstrated significant profit improvements, achieving an 18% increase in stock prices for 2024 and an 8% decrease entering 2025.
From an industry-wide perspective, electric power shares have broadly performed well over the last two yearsNotably, AI computing's boost to the power system extends beyond the bottom line; smart grid optimizations can analyze historical data and weather factors to enhance the accuracy of electricity demand forecasting while enabling the grid to better balance supply and demand. Additionally, substantial achievements have been made in integrating renewable energy, with predictions of wind and solar energy generation having been enhanced using AIThis enables more effective management of these intermittent energy sources, while storage optimization ensures charging and discharging strategies are most efficient to maximize renewable energy utilizationAI holds significant advantages in power supply management, network security, equipment management, and maintenance, while improving customer service and experience substantially.
In summary, AI’s applications within the electric power system are vast, encompassing grid operation optimization, enhancing renewable energy utilization, boosting network security, and elevating user servicesAI increasingly drives the power system towards a smarter, more efficient, and sustainable futureThe vigorous development of AI computing in China is similarly supported by its electric power system, highlighting that in the era of AI, electricity reigns supremeIn 2024, China retained its top position globally for electricity generation and installed capacity, greatly fulfilling the energy demands of AI advancements, and this symbiotic relationship bodes well for innovation and efficiency gains within electric power enterprises
Advertisements
Advertisements