Asia cloud computing, AI to make LNG the main power source

Last year, the Asia-Pacific region’s rapid growth in cloud computing, digitalisation and 5G usage helped total operational data capacity reach some 10.6 gigawatts, a new Wood Mackenzie report found.

Moreover, artificial intelligence (AI) will also drive increased demand for data centres in the region, prompting the need for more reliable power. These power-intensive data centres will also likely turn to gas-to-power for more electricity generation.

The report analysed four key Asian countries (Japan, South Korea, Thailand and Singapore) which already rely heavily on gas-fired power generation. Data centres could account for between 2 million tonnes per annum (mtpa) to 7 mtpa of extra gas demand across the four countries by 2030, the report said, with most of this demand being met by LNG.

Surprisingly, this demand could represent between 2 percent to 5 percent of global LNG imports. Global LNG trade is already substantial, reaching 404 million tonnes in 2023, up from 397 million tonnes the previous year, according to Shell LNG Outlook 2024.

Japan and South Korea

Both Japan and South Korea have a prominent role in the semiconductor and data centre value chain. These sectors will surge in demand to support the growth of AI, the report said. Based on assumptions from governments and local industry, combined power demand from data centres and semiconductor manufacturers could reach 37 terawatt hours (TWh) in Japan and 27 TWh in South Korea by 2030.

Japan’s use of gas to drive its data centres is based at least in part on its inability to jump start its nuclear power sector. The industry still faces considerable hurdles, particularly more cumbersome technical reviews, a reluctant court system, and lack of local government endorsement, compared to more favourable endorsement on the national level. This reluctance to fully embrace nuclear power is traced back to the 2011 Fukushima Daiichi nuclear disaster.

The South Korean government also favours nuclear power that in turn would help provide electricity for more data centres, but conventional nuclear power could take another decade or more to fully develop in the country, according to the report. Small modular reactors are also still several years away from large scale commercialisation in the country. As such, in both Japan and South Korea, LNG and gas-to-power will be vital for electricity generation to support intermittent solar and wind power generation. The report did note, however, that both countries would have preferred to rely more on renewables for data centre power sources to help hit their respective net zero goals instead of gas-to-power.

Assuming gas will account for 40 percent of both countries’ requirements, total LNG demand for data centres and semiconductor manufacturing could reach 3.7 mtpa by 2030, according to a mid-range Wood Mackenzie case scenario. That’s a 3.3 mtpa increase from 2024, representing some 3 percent of regional LNG demand.

Singapore

Association of Southeast Asian Nations (ASEAN) member Singapore has a decades-long data centre head start over other countries in the region. It’s seen rapid expansion of the industry since the early 2000s. The city state estimated that its data centres accounted for as much as 7 percent of total electricity consumption by 2020. Amid concern over data centre environmental impact, however, the government curtailed growth in the industry by putting a moratorium in place between 2019 and 2023.

Trying to assuage further concerns over economic impact from rampant data centre usage, the Singapore government in May introduced a comprehensive Green Data Centre Roadmap. This is intended to steer the industry towards sustainability, resource efficiency and innovative cooling solutions.

However, gas and LNG will dominate Singapore’s fuel mix until at least 2030, according to the Wood Mackenzie report.

This includes gas’ role in meeting most data centre power demand despite the roadmap’s focus on the adoption of greener energy sources. Consequently, future data centre market growth in Singapore will remain constrained with incremental gas demand limited to just 0.22 mtpa in the report’s base scenario, most of which will be LNG.

Thailand

Thailand, also an ASEAN member, and the bloc’s second largest economy, offers a different scenario. The country relies on gas, including domestic gas production in the Gulf of Thailand, pipeline gas from neighbouring Myanmar, and LNG imports, for 68 percent of its energy needs.

Moreover, unlike Singapore, which has controlled future data centre growth, over the past three years, the Thai government has made growth in the industry a priority. Its Thailand 4.0 initiative, which includes an unlimited eight year corporate income tax exemption with a potential five-year extension, is already spurring substantial investments in data centre infrastructure. Major investors on board so far include Microsoft, Google, Huawei, AWS ST Telemedia and NTT.

Gas reliance in Thailand will increase to help provide data centres with what the report called  “a source of continuous, dependable energy.” Data centres in Thailand will need around 6 TWh of electricity by 2030, or around 2.5 percent of total Thai power demand. If data centre projects materialise and investments continue, the high-end case could see gas demand in Thailand reach 0.9 mtpa (3 percent of total gas demand). In a lower case scenario, demand would be 0.2 mtpa, accounting for only 0.4 percent of total gas demand.

What the report doesn’t mention, however, is the possible environmental pushback that more gas usage from additional data centres and AI development in Asia could bring. In January, the International Energy Agency said in its forecast for energy usage over the next two years that electricity consumption associated with data centres, cryptocurrency and AI combined, represented nearly 2 percent of total global energy usage in 2022.  That amount could double in just four years by 2026, it added. This would be roughly equivalent to Japan’s total energy usage.

These warnings have led to calls for more transparency over all aspects of AI development, including energy usage, along with the need for AI to have a greener footprint going forward.