Gartner: By 2027, 40% of AI data centers will be constrained by power shortages

The rapid growth of GenAI energy consumption will exceed the ability of power companies to bear

Gartner predicts that artificial intelligence (AI) and generative artificial intelligence (GenAI) are causing electricity consumption to soar, with data center electricity consumption expected to grow by as much as 160% in the next two years . Gartner predicts that by 2027, 40% of existing AI data centers will be limited in their operations due to insufficient power supply.

“The number of new hyperscale data centers being built to implement generative AI is multiplying, overwhelming the supply of electricity, and even if utilities rapidly expand their own generation capacity, it will not be enough to meet this demand,” said Bob Johnson, research vice president at Gartner. “This in turn could disrupt energy supplies and cause energy shortages, limiting the number of new data centers built for generative AI and other purposes through 2026 and beyond.”

Gartner predicts that by 2027, the electricity consumption required to run new AI-optimized servers in data centers will reach 500 terawatt hours (TWh) per year, 2.6 times that of 2023 (see Figure 1).

Figure 1: Forecast of increase in electricity consumption of AI data centers from 2022 to 2027

Source: Gartner (November 2024)

“Large language models (LLMs) are fundamental to generative AI applications. To handle the massive amounts of data required to train and implement rapidly scaling LLMs, new large data centers are being planned,” Johnson said. “But short-term power shortages could persist for years because new transmission, distribution, and generation capacity could take years to come online and not solve immediate problems.”

In the near future, the number of new data centers and the growth of GenAI will be constrained by power supply. Gartner recommends that enterprises determine the risk that potential power shortages will pose to all products and services.

Electricity prices will rise

Gartner predicts that the impending power shortage will inevitably lead to higher electricity prices, which will also increase the operating costs of LLM.

“Large electricity users are working with major power producers to secure their own power sources over the long term and not be impacted by other grid demands,” Johnson said. “At the same time, data center operating power costs will increase significantly as operators use economic leverage to secure power demand. These costs will also be passed on to AI/GenAI product and service providers.”

Gartner recommends that companies evaluate future plans that anticipate rising electricity costs and negotiate long-term data center service contracts at reasonable electricity prices. Companies should also take the significant cost increases into account when planning new products and services, while looking for alternatives that consume less electricity.

Sustainable Development Goals will be affected

Short-term solutions to increase power will also have a negative impact on zero-carbon sustainability goals, as surging electricity demand forces suppliers to do whatever it takes to increase generation, in some cases even by allowing fossil fuel plants that were scheduled to retire to continue operating beyond their proposed closure dates.

“The reality is that generating the electricity needed to meet the greater power demands of data centers in the short term will result in an increase in CO2 emissions,” Johnson said. “This, in turn, will make it more difficult for data center operators and their customers to achieve aggressive sustainability goals related to CO2 emissions.”

Gartner believes that data centers need 24/7 power supply, which cannot be met by renewable energy sources such as wind or solar without some form of alternative power supply during the period when power is not generated. Only hydroelectric power plants, fossil fuel power plants or nuclear power plants can provide reliable 24/7 power supply. In the long run, new technologies will emerge to improve battery storage (such as sodium-ion batteries) or provide clean energy (such as small nuclear reactors), which will help companies achieve sustainable development goals.

Gartner recommends that companies reassess their sustainability goals related to CO2 emissions based on data center demand and electricity sources in the coming years. When developing GenAI applications, companies should focus on minimizing computing power consumption and explore the feasibility of other options such as edge computing and smaller language models.

Gartner clients can learn more in "Emerging Technologies: Power Shortages Will Limit Growth and Implementation of GenAI."