Global energy storage system solution supplier with 15 years manufacturing experience. Chinese
Language
Global energy storage system solution supplier with 15 years manufacturing experience. Chinese
In today's rapidly advancing AI technology landscape, a concerning reality is emerging: the power required for a single ChatGPT query is nearly 10 times that of a Google search.
This significant gap not only highlights the fundamental difference in energy consumption between AI technologies and traditional internet services but also signals a profound shift in the global energy consumption pattern.
Recently, renowned consulting firm Gartner issued a warning in its latest report, predicting that by 2027, 40% of existing AI data centers will face operational difficulties due to insufficient power supply. This forecast underscores the growing tension between AI development and energy supply.
At the same time, research from international investment bank Goldman Sachs provides a similar outlook: by 2030, global data center electricity demand will surge by 160%. This has sparked widespread concern regarding energy supply, infrastructure development, and environmental impact.
Chart | Gartner's Forecast: Additional Energy Consumption from New AI Servers in AI Data Centers Each Year (Source: Gartner)
Recently, tech giants like Google, Microsoft, Amazon, and Meta have been actively investing in nuclear power facilities. One of the reasons for this is their concern that the immense energy demand of AI data centers in the future may not be met.
Historically, the energy demand of data centers has shown remarkable stability. From 2015 to 2019, despite nearly doubling the workload of data centers, their annual electricity consumption remained relatively stable at around 200 terawatt-hours.
This stability was largely due to continuous improvements in energy efficiency within data centers. However, this situation underwent a fundamental shift after 2020.
Gartner analyst Bob Johnson noted, "The construction of next-generation, hyperscale data centers is creating enormous electricity demands that will outpace the ability of utility providers to scale up supply. Particularly in the realm of processing and training large models, the required computational resources and energy consumption have reached unprecedented levels."
Currently, global data centers account for 1-2% of total electricity consumption, but it is projected that by 2030, this share will rise to 3-4%, with this growth being especially prominent in developed countries.
In particular, in the United States, it is predicted that by 2030, the electricity consumption of data centers will increase from the current 3% to 8%, driving U.S. electricity demand to grow at its fastest rate in nearly 25 years.
Chart | Goldman Sachs Forecasts Energy Demand for Data Centers (Source: Goldman Sachs)
To address this challenge, U.S. utility companies will need to invest approximately $50 billion in new power generation capacity specifically for data centers.
Additionally, by 2030, the increased electricity demand from data centers alone will result in a daily increase of about 3.3 billion cubic feet of natural gas demand, which will require the construction of new pipeline infrastructure.
Goldman Sachs notes that the situation in Europe is even more complex. As a major hub for global data centers, 15% of data centers are located in Europe. By 2030, the energy demand of these data centers will be equivalent to the total electricity consumption of Portugal, Greece, and the Netherlands combined.
Given that Europe has the oldest electricity grid systems in the world, the region will need to invest nearly €800 billion over the next decade to upgrade its transmission and distribution systems, as well as approximately €850 billion in the development of renewable energy sources such as solar, onshore wind, and offshore wind power to meet the energy needs of new data centers.
Chart | Average Age of Power Grids in Various Regions and China (Source: Goldman Sachs)
What's even more concerning is that this surge in electricity demand will directly impact electricity prices. Research indicates that large data center operators are negotiating with major power producers to secure long-term, stable electricity supplies independent of other grid demands.
This competition will inevitably drive up electricity prices, and these costs will ultimately be passed on to users of AI products and services.
As a result, experts recommend that organizations prepare in advance for rising electricity costs and strive to sign long-term data center service contracts at reasonable prices.
The environmental impact is also worrying. It is expected that by 2030, the carbon emissions from data centers may more than double compared to 2022, presenting a new challenge to global emission reduction targets.
According to Goldman Sachs, the "social cost" of the increase in carbon emissions from data centers alone will amount to $125 billion to $140 billion (present value).
Gartner estimates that by 2027, the electricity demand for running AI-optimized servers will reach 500 terawatt-hours per year, 2.6 times the level in 2023.
In the short term, to meet the surging electricity demand, some fossil fuel power plants that were originally scheduled for decommissioning may have to extend their operational life, further exacerbating environmental pressures.
Data centers require 24-hour uninterrupted power, and currently, they must rely on hydroelectric, fossil fuel, or nuclear power plants to provide such stable electricity supply.
While renewable energy sources like wind and solar are environmentally friendly, without supporting energy storage systems, they are difficult to rely on to meet the continuous power demands of data centers.
Chart | Changes in Data Center Load and Energy Consumption Over the Past Nine Years (Source: Goldman Sachs)
To address these challenges, the industry is exploring various solutions. Some companies are increasing investments in renewable energy and actively promoting the commercialization of new nuclear power technologies.
Tech companies are also exploring innovative methods to improve energy efficiency. In the long run, the development of new battery storage technologies or clean energy technologies (such as small nuclear reactors) may provide new solutions.
It is worth mentioning that AI technology itself could contribute to solutions by accelerating innovation in fields such as healthcare, agriculture, and education, as well as improving energy efficiency.
Finally, both companies' research reports suggest that businesses should fully consider the potential risks of power shortages when formulating AI development strategies, assess the impact of rising power costs in the future, and actively seek alternative solutions.
Promising solutions include using edge computing technologies, adopting smaller large models, and prioritizing computational efficiency when developing generative AI applications.
Clearly, the development of AI technology is reshaping the global energy landscape. Balancing technological innovation, energy security, and environmental protection will be a significant challenge that the global tech and energy industries will face together in the future. (Article republished from DeepTech
Blue Carbon Battery, which officially entered in Rizhao in 2009, is one of the best suppliers of photovoltaic product research development and micro-energy storage system solutions in the world.
READ MORE >>
