Data centers and small reactors could change Asia's nuclear dynamic

Japan Times

3 days ago

Notice how hot your cellphone gets when it multitasks on a steamy day? It doesn't take much to make me put my keitai down and worry about it overheating.
Multiply that heat by, say, infinity, and you've got the biggest problem that data centers face in an increasingly digital world. An estimated 402.74 million terabytes of data are created each day and storing and processing all that information creates virtual volcanoes. The rise of artificial intelligence is compounding the problem.
In a much-cited report released earlier this year, the International Energy Agency (IEA) estimated that electricity demand from data centers worldwide will more than double by 2030 to around 945 terawatt-hours, an amount that exceeds Japan's current entire electricity consumption. By 2035, global data center electricity consumption will increase again by one-third to around 1,200 TWh. (Goldman Sachs reached roughly similar conclusions in reports issued last year.)
Data center electricity consumption has grown 12% a year since 2017 — more than four times the rate of total electricity consumption — and last year accounted for around 1.5% of the world's electricity consumption in 2024, or 415 TWh. The U.S. accounted for 45% of that amount, followed by China (25%) and Europe (15%).
It's estimated that a typical AI-focused data center consumes as much electricity as 100,000 homes and the largest ones currently under construction will consume 20 times that amount, or the equivalent of 2 million households.
A huge part of that demand comes from the cooling systems critical to their operation. New technology such as graphics processing units, needed for increasingly complex computation, consume more power and generate more heat than the central processing units that are more widely used today. It's estimated that cooling systems can account for as much as 40% of a data center's power consumption.
The IEA sources 85% of global data center power consumption — and hence data centers — to the U.S., China and Europe. In Japan, data centers account for less than 20 TWh of electricity consumption (about 2% of the total, roughly equivalent to that of Europe).
That will change. The transition to Society 5.0, with its deep integration of digital services into daily life, will accelerate demand for computing power. In addition, the Japanese government has made data centers a core component of efforts to increase foreign direct investment in the country to ¥100 trillion by the end of 2030.
Japan is already scheduled to add more than 500MW to power its data centers, bringing total capacity for those facilities to 1.5 GW. Consistent with those plans, the IEA concluded that data centers will be responsible for more than half the growth in electricity demand in Japan.
The rest of the world accounts for about 10% of total data center electricity generation, with Southeast Asia and India prominent. In many ways, Southeast Asia may be the most interesting — and concerning — place to watch.
A BCG report last year anticipated that data center capacity in Southeast Asia would triple by 2030, reaching between 5.2 GW and 6.5 GW, which would make it the world's third hub for such services, behind the U.S. and China. This will, reports the IEA, double electricity demand in the region. As one example, the agency forecast that data centers would be responsible for as much as 20% of the increase in Malaysia's national power demand by 2030.
Ambitious politicians, such as Thaksin Shinawatra, Thailand's eminence grise, hope to emulate Japan and use that demand to entice foreign investment. 'We want to be competitive in data centers and AI,' Thaksin said earlier this year. Bangkok has already reportedly secured billions of dollars in investment commitments by technology giants like Amazon and Alibaba.
The big question is where they will get their juice. Rising demand is pushing tech giants such as Google, Microsoft and Amazon to invest in nuclear power. This shift, from governments to big companies, as the primary drivers of nuclear power development, has powerful implications that we have been slow to consider.
Big nuclear projects remain problematic, with government policies and regulations inhibiting — if not blocking — the sector's development. There is only one nuclear power plant in Southeast Asia — in the Philippines, built 40 years ago but never commissioned. It has never produced a single watt of energy.
Yet, in a report issued late last year, the IEA concluded that nuclear power would become part of the region's energy mix by 2035, with installed nuclear capacity ranging from 6 GW to 13 GW by 2050, depending on the scenario.
The development of small modular reactors is the wild card. SMRs have a capacity of under 300 MW (small, most conventional reactors produce about 1,000 MW) and their components can be mass-produced and assembled on site (modular). They are also considerably cheaper than traditional nuclear power plants: They cost about $2 billion while conventional reactors can cost more than $10 billion. Thus far, only Russia and China have commercially operable reactors, although the U.S. and Japan are working on the technology.
According to the IEA, there are plans to build as much as 25 GW of SMR capacity to supply data centers worldwide, almost all in the U.S., with the first projects expected to go online after 2030. But does the readiness of tech giants to acquire their own nuclear capability change the regional dynamic?
SMRs are intended to sidestep the big issues surrounding nuclear power, namely the huge costs required to build the plant and the accompanying grid as well as looming and rightly concerning safety and security issues. SMRs can be installed into an existing grid or remotely off-grid, reducing or even eliminating the infrastructure concerns. They're also thought to be safer and more secure.
While those reactors will still be subject to national regulatory frameworks, it seems obvious to me that moving the locus of decision-making away from governments to companies could accelerate the adoption process. Especially when those companies are pursuing business objectives that align with national economic goals — promoting investment and the development of high-tech industries. Companies can move more quickly than governments. Their readiness to assume regulatory and financial burdens helps shift the decision-making dynamic.
I checked in with Carl Baker, my colleague at Pacific Forum who has been running programs that explore regional thinking about nuclear energy for nearly two decades. He agreed that 'the potential availability of SMRs has had an impact on thinking about the viability of nuclear technology, especially in Southeast Asia.' But, he added 'the conversation regarding the introduction of nuclear energy as a source of electrical power is still in its early stages and has not gotten into the specific issues related to governance and other regulatory considerations.'
He is less sanguine than I about the impact of the private sector's enthusiasm for SMRs. The idea that tech giants can change the prevailing mindset 'is not really an issue in countries that have not developed the national regulatory framework for use of nuclear power as an energy source.' Moreover, 'conglomerates are not going to be able to avoid the bottlenecks and roadblocks any more than large power companies have been.'
Finally, Baker said, 'the impetus for accommodating specialized applications is coming primarily from the U.S at this point' and regulators there have been pretty reluctant to let go of overly stringent safety requirements.
He is probably right. But in a recent conversation in Southeast Asia about nuclear power and accompanying concerns, local participants approached the problem from the traditional perspective — big nukes, national infrastructure and government bottlenecks. Those are powerful forces but I can't help but wonder if that paradigm is ready for revision. After all, things are heating up.
Brad Glosserman is a senior adviser at Pacific Forum and the author of "Peak Japan." His upcoming book on the geopolitics of high-tech is expected to be released by Hurst Publishers this fall.