In recent years, the rise of the artificial intelligence (AI) industry has driven the vigorous development of Taiwan’s semiconductor manufacturing sector, and Taiwan’s growth in electricity demand has exceeded expectations. AI is built on computing power; computing power is national power, and computing power is effectively equivalent to electric power. Whether Taiwan actually has sufficient electricity to support the needs has already triggered alarm in industry.
When asked by the media，the Minister of Economic Affairs’ assurance that Taiwan would not experience power shortages before 2034, AI “godfather” Jensen Huang offered a thought-provoking word: “Maybe.” It signifies that while the plan sounds feasible, there is widespread skepticism in industry as to whether the government’s plan can be implemented on schedule, and thus a generally cautious attitude about whether Taiwan will face power shortages. In a Facebook post, Huang Ching-yung, founder of Digitimes, recounted a one-on-one conversation with Jensen Huang, during which Huang said, “Once manufacturers start having doubts about the infrastructure of water and electricity supply, they may turn to ASEAN and South Asia, or to China for production.” He observed the seriousness in Huang’s expression: “Has Taiwan seized a once-in-a-lifetime opportunity, only to miss it because of inadequate infrastructure?” Jensen Huang’s concerns must surely reflect what he has been hearing from many Taiwanese business leaders in his conversations with them.
The government should long have been aware of Taiwan’s electricity predicament. On March 21, 2026, when President Lai Ching-te attended a public event, he mentioned that the Ministry of Economic Affairs had carefully evaluated and concluded that the Second and Third Nuclear Power Plants of Taiwan Power Company (Taipower) meet the conditions for restarting operations. Under the premise of legal permissibility, he agreed to proceed with life-extension work for both plants. The Democratic Progressive Party (DPP) administration’s energy policy has thus embarked on a “road back to nuclear power.” The saying “It does not matter if it lasts forever; what matters is that we once had it” is perhaps the most fitting description of Lai Ching-te’s attitude toward the previous “nuclear-free homeland” policy of government.
There are four concrete tasks on this “road back to nuclear power,” listed in order of the time required for completion: extending the operating license of existing nuclear stations; bringing the Fourth Nuclear Power Station into commercial operation; building new large convention nuclear units at existing sites; and introducing small modular reactors (SMRs). These four tasks can proceed in parallel.
Life extension or license renewal for nuclear power stations is a common measure in virtually all countries adopting nuclear power and is widely recognized as the most effective tool for reducing carbon emissions. In 2026 there are 415 operating nuclear units, globally, of which 184 have been in operation for more than 40 years and 40 for more than 50 years. In the United States, 22 nuclear units have already obtained 80-year operating licenses, and dozens of others are under review. Taiwan’s three nuclear power stations of six units have only operated for 40 years; once the relevant procedures are completed, they can of course continue to be used. Life-extension efforts at the Second and Third Nuclear Power Station are currently under way, but whether they can generate power on schedule still depends on resolving several detailed issues. The units at the First Nuclear Power Station remain intact and are available for life extension.
The Fourth Nuclear Power Station is currently mothballed, but it remains listed as an asset of Taipower. For the sake of its creditworthiness, Taipower must properly maintain the facility. It is beyond doubt that, with appropriate additional investment, the station can be brought into commercial operation. Investment in the Fourth Nuclear Power Station has been about NT$300 billion; an estimated of additional NT$30–100 billion is required to complete the project. Based on recent experience in Europe and the United States (such as Vogtle Units 3 and 4 in the United States and Hinkley Point C in the United Kingdom), construction costs have soared to more than US$8,000 to US$10,000 per kilowatt. Even in Japan, South Korea, or mainland China—where plant construction is more efficient and supply chains are more complete—costs still range from US$4,000 to US$5,000 per kilowatt, equivalent to roughly NT$420 billion to NT$450 billion for a plant with the capacity of the Fourth Nuclear Power Station. If one also factors in Taiwan’s unique requirements for environmental impact assessments, local compensation funds, and interest accrued over construction periods exceeding 10 years, total expenditures will easily exceed NT$500–600 billion.
According to data published on the International Atomic Energy Agency website, there are currently 72 units under construction in 16 countries worldwide, of which only two are small modular reactors. Conventional large-scale nuclear reactors remain dominant in the market. There are several choices of technologically mature units with extensive construction and operating experience available. All four of Taiwan’s nuclear power station sites have sufficient space to accommodate new units. In 2010, Taipower completed an internal feasibility study to construct Nuclear Units 9 and 10. Taipower had begun the process of selecting a consulting company to prepare the formal feasibility assessment report at the time, but the plan was withdrawn after the Fukushima accident.
There are a great many conceptual designs for small modular reactors, but only a few have entered licensing review, and even fewer have begun construction. It may not be until around 2032 that any units achieve commercial operation. The following points must be clearly understood with respect to SMRs: their generation costs will be significantly higher than those of traditional large nuclear unit; SMRs are by no means absolutely safe, but only relatively safer compared with large reactors; and SMR operation will still produce nuclear waste. It is unrealistic for Taiwan to design its own SMRs, and there is no stock available for market purchase at present. Given Taiwan’s current situation, SMRs are not a timely solution.
Taiwan’s road back to nuclear power requires innovative approaches. Taipower is a state-owned enterprise whose budgets must be approved by the Legislative Yuan; obstruction by a single legislator can lead to several months of delay. Implementation of public budgets is constrained by the Government Procurement Act, resulting in lengthy procedures. If any tender is disputed, it can hold up the entire schedule. Execution of works tendered under the Procurement Act lacks flexibility, and public entities have limited bargaining chips at their disposal. Differences in contract interpretation among Taipower project officers, contractors, and the Nuclear Safety Commission can lead to irresolvable disagreements, all of which cause delays of the project.
The innovative approach is to adopt a “state-owned, privately operated” model for nuclear power station operations. Concretely, follow the procedure of the Procurement Act, the nuclear power stations of Taipower would be transferred to private companies for operation under an ROT (Renovate–Operate–Transfer) arrangement. The successful bidder would pay Taipower annual concession fees in accordance with the bid documents. The concession fees could be calculated using a dual-track system based on capacity and total power generated annually. The bidder would bear the costs required for life-extension and improvement works at the Stations. Through contractual arrangements, the bidder would commission Taipower’s nuclear power division to carry out the plants’ day-to-day operation and maintenance. The bidder would be required to seek cooperation with international nuclear power utility that have strong track records in nuclear plant operation—ideally through joint ventures, but consultancy arrangements would also be acceptable.
All electricity generated by the nuclear plants would belong to the successful bidder. The options for selling this electricity would be: (1) supplying zero-carbon electricity directly to end-users; (2) selling electricity to Taipower under contract, following the model of independent power producers (IPPs); or (3) selling electricity to Taipower under contract following the cogeneration model of Taipower. All three modes would require Taipower to provide wheeling services. Taipower already has precedents for all three models.
The bidder’s profits would derive from the option to supply zero-carbon electricity to large manufacturing enterprises at a higher price. Although such tariffs would be higher than Taipower’s retail price, nuclear stations’ abundant output would allow companies to adjust the proportion of nuclear power they purchase according to their products’ competitiveness needs, enabling certain manufacturers to achieve genuinely zero-carbon products.
Under the state-owned, privately operated model, the concession fees received by Taipower could be booked directly as negative costs in its average generation cost calculation, thereby lowering the company’s average cost of generation without affecting customer interests. Only under a state-owned, privately operated model is it possible to supply nuclear power to firms that urgently need zero-carbon electricity to enhance their competitiveness, thereby highlighting the actual value of nuclear power through market-based supply-and-demand principles. The beneficiaries would be Taipower’s customers, industrial enterprises, investors, and Taipower itself. If the ROT life extension model for the Second and Third Nuclear Power Stations can be successfully implemented, the same model could later be applied to the commercial operation of the Fourth Nuclear Power Station.
The entrustment of public facilities to private operation is legally grounded, backed by precedents, and guided by clear statutory procedures; it does not require approval from the Legislative Yuan. Internationally, there are also precedents where the owner and the operator of a nuclear power station are different companies—for instance, nuclear power plants in the United Kingdom are operated by France's EDF. The selling of the power generated directly to customers by the winning bidder is legally compliant and does not require new legislation.
Investing in carbon-free nuclear power offers substantial profits. This should attract the attention of investors, and even international capital. The next step is to see whether enterprises are willing to purchase carbon-free nuclear power at an appropriate price. If, prior to the tendering process, businesses can make concrete commitments expressing their willingness to purchase nuclear electricity, the winning bidder can negotiate prices and sign contracts with these businesses after the contract is awarded. This should allow investors to secure bank financing.
The greatest challenge to this concept is whether investors have doubts regarding the continuity of government policies. This will require the determination of the government agency to boost investor confidence.