[Column Series] Key Issues to Address in Japan's Strategic Energy Plan (No.2) Growth of AI Will Not Negatively Impact Electricity Supply and DemandEven Help to Achieve 100% Renewable Energy

In discussions about the Japanese government’s Strategic Energy Plan, the possibility that growing use of artificial intelligence (AI) will drive up electricity demand has drawn significant attention. This scenario assumes an expansion in data centers for performing the information (data) processing required for AI, along with rising production of the semiconductors needed for that processing. At a meeting of the Strategic Policy Subcommittee  (June 6, 2024) that is discussing the Strategic Energy Plan, forecasts of electricity demand were presented by five research institutes (Chart 1).

Chart 1: Forecasts of electricity demand in Japan

According to the Agency for Natural Resources and Energy, which put together these forecasts, three of the five research institutes (CRIEPI, RITE, and Deloitte) explicitly considered the likelihood of growing electricity demand for data centers and semiconductor production. The forecasts of total electricity demand in FY (Fiscal Year) 2050 vary widely between the five institutes. At the low end, demand will be much the same as in FY2022, while at the high end, demand will be 40% or so higher. This variation reflects the difficulty of predicting the future.

A look back over the past suggests that a substantial expansion in AI-related products and services is unlikely to drive up electricity demand by any significant amount. A useful comparison is the growth of the Internet in Japan, which exploded in the latter 1990s. For over 20 years, the Internet has dramatically reshaped both the industrial structure of the world and human lifestyles. Internet and AI are indistinguishable to a large extent, and in any case, they are both powered by data centers and semiconductors. Both are novel forces of transformation in human society through the emergence of information technologies (IT).

In Japan, the number of individual Internet users grew very fast between 2000 and 2005 (Chart 2). After accelerating from 9.2% of the population in 1997 to 37.1% in 2000, it had quickly reached 70.8% by 2005. And usage has continued to rise steadily to the point that most people today use a wide variety of Internet-based services every day. Obviously, this massive expansion in the use of personal computers, smartphones, and other devices that connect to the Internet has driven up the demand for data centers and semiconductors enormously.

Chart 2: Rate of Internet usage in Japan (individuals)

On the other hand, domestic electricity consumption (by year) indicates a gradual rise from 1990 to 2007 (Chart 3). Electricity consumption grew in line with the progressive electrification of homes and industrial facilities. However, from 2000 to 2005, when Internet users were growing rapidly, electricity consumption only increased by 5%. Since electricity consumption peaked in 2007 (the year before the financial crisis of 2008), it has continually declined, despite a huge growth in the use of Internet-related products and services. This has been due, in large part, to energy efficiency efforts.

Chart 3: Electricity consumption in Japan

Despite the massive growth of the Internet, Japan’s total electricity consumption has fallen. This is especially remarkable when considering that most of the population now uses the Internet for long periods of time every day, both for work and private use. Even as the number of Internet-connected devices and the number of users of a vast range of services continues growing, electricity supply and demand has remained largely unaffected. The long-term shutdown of large-scale power plants has impacted supply much more significantly.

There are several possible reasons why Internet expansion has not notably impacted domestic electricity demand. The first is that the technologies on which the Internet is built have advanced. For example, the performance of semiconductors has improved at a stunning pace, doubling every year and a half to two years, or by a factor of 100 in a decade. The processing of massive volumes of data can now be done with a tiny fraction of the power required previously. Secondly, the Internet has made many business processes far more efficient. If companies shorten their business operation hours, for example, they can save on energy for air conditioning, lighting, and equipment control. Thirdly, information processing for the Internet does not necessarily need to be done in Japan, i.e., data can be processed at data centers overseas. Similarly, as AI-related products and services spread, the high cost of electricity and Japan’s high level of thermal power generation, which accelerates climate change, may lead to most businesses deciding to use overseas data centers.

Essentially, it is hard to expect any technologies consuming huge amounts of electricity to spread worldwide. New technologies and services will only take off if there is a good balance between cost and benefit. As AI gets more and more established in the years ahead, software to enable highly efficient energy management will also evolve, along with advances in the efficiency of semiconductors and other hardware, thereby reducing the power consumption for intensive information processing even further. One of the reasons that Google and other IT industry leaders continue to make pioneering efforts in the energy sector is that efficient energy use is vital for expanding their new technologies and services. Naturally, avoiding excessive cost increases is also important to ensure the continuity of businesses.

Even if an age of massive power consumption does arrive due to AI, the ability to supply electricity at almost zero additional generation cost will be essential. This prerequisite can only be met by using renewable energy sources such as solar, wind, hydro and geothermal, requiring no fuels. In this sense, the widespread adoption of AI could pave the way to 100% renewable electricity across the world.

In terms of electricity supply and demand in Japan, the growing use of AI cannot be a reason for expanding nuclear or maintaining coal-fired power plants. Although some hopes for nuclear power generation based on fusion technologies have been expressed in recent times, this technology is not expected to be commercially viable until at least the 2040s. This would be too late to meet any rise in electricity demand due to AI growth. It is also unclear whether the cost of nuclear power generation can fall low enough to compete with renewable energy.

There is also a risk that over-estimating the demand for electricity due to AI growth may result in high-cost power generation facilities becoming economically unsustainable, ending up as stranded assets. Consequently, this could put pressure on the power generation business, even threatening Japan’s major electric power companies. On the other hand, if the expansion of renewables is delayed by policy to avoid the risk of stranded assets, Japan could lose out in the global push to decarbonization. Neither situation is desirable for Japan.

Japan should now be aiming to develop a system capable of supplying large quantities of electricity at the lowest possible cost by making maximum use of fuel-free renewable energy sources. If AI does become indispensable to future industry and lifestyles, inexpensive electricity will be essential. Given the urgency of climate change mitigation, expanding the supply of electricity produced with minimal greenhouse gas emissions and environmental impact should be a top priority. Focusing on nuclear and coal-fired power generation on the assumption of an excessive rise in electricity demand due to AI is likely to be a serious misstep in Japan’s energy strategy.