Chips need current as AI pushes Thailand, Southeast Asia toward small reactors

Southeast Asia’s race to attract artificial intelligence investment and data centres is forcing governments to confront an awkward question: where will the electricity come from? For Thailand, one increasingly serious answer is nuclear power.

Bangkok has included two small modular reactors, or SMRs, with combined capacity of 600MW towards the end of its draft Power Development Plan 2026, marking a significant shift for a country that has discussed nuclear power for decades without building a commercial reactor.

In mid-September on the sidelines of the Gastech conference in Bangkok, the Electricity Generating Authority of Thailand signed an agreement with the US Trade and Development Agency to fund a feasibility study of American SMR technology capable of supplying roughly the same 600MW of electricity. The immediate catalyst is not just decarbonisation, it is electricity demand.

Sanjeev Gupta, senior partner at EY-Parthenon and moderator of a Gastech discussion on nuclear power and artificial intelligence, mentioned that the rapid expansion of AI and data centres was placing growing pressure on existing power systems and helping revive interest in nuclear energy.

The return of nuclear to Southeast Asia’s policy debate comes as electricity consumption across the region is already expanding rapidly. The International Energy Agency estimates electricity demand has grown by about 6% a year since 2015, against 4% growth in overall energy demand.

Over the next decade, the increase in Southeast Asian electricity consumption alone will be equivalent to Japan’s entire present electricity generation, the IEA estimates. Data centres are becoming an increasingly important source of that demand, alongside cooling, manufacturing and wider electrification.

The region is already responsible for about 10% of data-centre electricity consumption in Asia-Pacific, according to the agency, with demand becoming increasingly concentrated around the emerging digital hubs of Singapore, Malaysia, Thailand and Indonesia. That is changing the economics of nuclear power.

Unlike much existing computing infrastructure, large AI data centres can require hundreds of megawatts — and increasingly gigawatts — of electricity around the clock. Solar and wind can supply cheap low-carbon electricity but require storage, grid investment or complementary generation to provide continuous power.

SMRs promise a different solution: smaller nuclear plants that supporters argue could eventually be manufactured more systematically and deployed incrementally rather than requiring the huge upfront investment associated with conventional multi-gigawatt reactors.

That proposition has begun to attract Asian policymakers precisely as hyperscale computing companies are looking for large quantities of reliable low-carbon electricity.

Nateekool Kriangchaiporn, a senior nuclear engineer at EGAT, told the Gastech panel that the Thai utility was planning for the country’s first SMR project and hoped to make substantial progress during the next decade.

She said data centres and AI had become a new issue in Thailand’s electricity planning, alongside the need to accommodate much greater quantities of renewable generation. Nuclear could potentially provide dependable generation while helping stabilise a grid increasingly supplied by intermittent solar and wind power.

Energy security is adding urgency to the calculation. Thailand currently gets about 30% of its electricity from imported LNG, according to state-controlled oil and gas producer PTTEP. The company warned at Gastech that the figure could rise sharply as domestic gas production declines and pipeline supplies from Myanmar become less dependable.

Every $3 per million British thermal units increase in LNG prices could raise Thai electricity prices by roughly 5%, PTTEP said, illustrating the exposure of an increasingly import-dependent electricity system to volatile international fuel markets.

Thailand is far from alone in reconsidering nuclear energy. Singapore, whose lack of domestic energy resources has long made security of supply a strategic concern, will begin an International Atomic Energy Agency review in 2027 to assess whether it has the institutions and expertise necessary to make an informed decision on deploying advanced nuclear technologies.

The renewed enthusiasm comes with a substantial caveat, however: the SMR industry remains immature. Thomas Hardy, deputy director of USTDA, told the Gastech panel there were roughly 120 technologies that could potentially be categorised as SMRs or microreactors, but warned that “not every one of those will succeed”.

Governments considering nuclear power must also build far more than a reactor. They need independent regulators, engineers, emergency planning, nuclear security arrangements, fuel strategies and systems to handle radioactive waste. The IAEA identifies 19 separate infrastructure issues that countries should address when developing a nuclear power programme.

Kriangchaiporn said public confidence could be one of Thailand’s most difficult challenges. People wanted answers to basic questions, she said, including how radioactive waste would be managed and “if something goes wrong, what happens?” Building that trust “needs time”, she added, alongside information that the public can understand rather than purely technical explanations.

The experience of the US also suggests why AI could make this nuclear cycle different from previous abortive “renaissances”. Jason Lanclos, director of energy development and planning for Louisiana, said that only two or three years ago the state expected advanced manufacturing to be the principal driver of future electricity growth. AI had “completely turned those assumptions on their head”, he said.

Louisiana was now facing an “unprecedented amount of power demand” from data centres alongside chemicals, advanced manufacturing and other energy-intensive industries. One large data-centre development alone was expected to require roughly 2.7GW to 3.2GW, according to Lanclos.

The result is an unusual convergence between technology companies, governments and a nuclear industry that has spent years searching for customers capable of underpinning expensive new projects with long-term electricity demand.

For Southeast Asia, the starting point is much further back. Thailand has no commercial nuclear power plant, and its proposed SMRs remain at the planning and feasibility stage. Singapore is still evaluating whether nuclear should be used at all. But the electricity problem they are trying to solve is arriving faster than the reactors. As the regional Newsbase Horizon analyst Mark Buckton said at Gastech, "Southeast Asia is doing all the right things on LNG infrastructure, and renewables are booming across the region, but that won't be enough on its own” adding “you have a population of just over 700mn growing at almost 0.7% a year, plus industrialisation and data centres coming on top of that. The region needs firm, zero-carbon baseload, and that means nuclear has to be part of the mix.”

The IEA expects Southeast Asian power demand to grow by an average 5.3% annually between 2026 and 2030. At the same time, governments are already attempting to decarbonise electricity systems while reducing vulnerability to imported fossil fuels.

That combination has put an old technology back into a very new debate.

Because of this, the question now facing Southeast Asia is no longer simply whether nuclear power is acceptable. The rise of AI is increasingly forcing policymakers to ask whether grids built for the previous industrial era can supply the enormous quantities of continuous electricity required by the next one — and whether small reactors can be built quickly and cheaply enough to matter.