TW-G-05: Taiwan Energy Policy, Nuclear Politics, and the Security-of-Supply Question — From the 1978 Chinshan Reactor Start-Up through the Nuclear-Free-Homeland Phase-Out to the Post-2024 Reversal Debate (1978–2026)

Status: DRAFTWords: 9,989

1. Key Takeaways

Taiwan is one of the most energy-import-dependent significant economies in the world — approximately 97–98 per cent of its primary energy supply is imported [TBD-VERIFY: Energy Administration figures consistently cite 97.5–98 per cent across 2015–2024 editions] — and this structural exposure converts every energy-policy choice into a national-security choice. Coal arrives from Australia and Indonesia, LNG from Qatar, Australia, and the United States, and oil from the Middle East, all across sea lanes that the People's Liberation Army has rehearsed interdicting since the 2022 post-Pelosi exercises (TW-C-06; TW-F-04). The most-cited single statistic in the post-2022 debate is natural-gas storage: roughly 11 days of LNG stocks in summer conditions [TBD-VERIFY: figures cited range from 8 to 14 days depending on season and source], against approximately 39 days of coal and around five months of oil. In a blockade scenario the gas-fired half of the electricity system is the first to fail, which is why energy policy migrated after 2022 from the economic-affairs portfolio into the defence-resilience conversation.
The demand side is dominated by the semiconductor industry whose strategic value is Taiwan's principal external shield — producing the distinctive Taiwanese paradox that the same sector which deters attack also deepens the energy vulnerability. TSMC alone consumed an estimated 7.5–9 per cent of Taiwan's total electricity by 2022–2023, with projections that the figure could reach 12–15 per cent by 2030 as 2nm-and-below nodes and AI-driven capacity expansion come online [TBD-VERIFY: TSMC ESG-report and Taipower figures]. The broader electronics ecosystem (TW-G-01, TW-G-02) pushes the industrial share of consumption above half. Every credible account of Taiwan's energy future is therefore also an account of whether the silicon shield (TW-N-01) can be powered — a framing TSMC itself reinforced through its RE100 renewable-procurement commitments and its quiet signalling about grid-stability requirements for advanced fabs.
The six-reactor nuclear fleet (Chinshan 1978/1979, Kuosheng 1981/1983, Maanshan 1984/1985) was a KMT developmental-state project — built by the state monopoly Taipower under authoritarian conditions, with no meaningful siting consent — and the anti-nuclear movement that arose against it became a foundational strand of Taiwanese democratic identity. Nuclear power supplied roughly half of Taiwan's electricity at its mid-1980s peak [TBD-VERIFY: peak-share figure circa 1985 cited between 41 and 52 per cent]. The storage of low-level nuclear waste on Orchid Island (Lanyu) from 1982, on the land of the Tao (Yami) indigenous people who were told the facility was a fish cannery [TBD-VERIFY: the "cannery" account is widely reported but contested in detail], became the movement's founding injustice. The DPP wrote a nuclear-free-homeland (feihe jiayuan, 非核家園) plank into its charter in the late 1980s: opposition to nuclear power is constitutive of the party, not incidental to it.
The Fourth Nuclear Power Plant at Lungmen is the single most expensive piece of unused infrastructure in Taiwanese history and the through-line of four decades of energy politics. Proposed in 1980, budgeted in the early 1990s, begun in 1999, halted by the first Chen Shui-bian government in October 2000, restarted in February 2001 after a constitutional crisis and Judicial Yuan Interpretation No. 520, and finally mothballed in April 2014 under the pressure of Lin Yi-hsiung's hunger strike and mass demonstrations — Lungmen absorbed more than NT$280 billion [TBD-VERIFY: cumulative-cost figures range NT$280–300 billion] without generating a single commercial watt. Its 2021 restart referendum failed. Lungmen is to Taiwanese energy politics what the 228 Incident is to identity politics: the reference event through which each side reads the other's bad faith.
Fukushima (March 2011) transformed Taiwanese nuclear politics from a partisan issue into a mass-civic one, and the 2013–2014 mobilisation peak — 200,000-plus marchers in March 2013, the April 2014 Lin Yi-hsiung hunger strike weeks after the Sunflower occupation ended — forced a KMT government to mothball Lungmen. Taiwan's reactors sit, like Japan's, on a seismically active coast; two of the three plants are within 30 kilometres of greater Taipei's evacuation-impossible millions. The Ma government's counter-move — daring the opposition to a referendum under the high-threshold "birdcage" Referendum Act — never came to a vote. The episode established the pattern that has governed energy policy since: direct-democracy instruments wielded tactically by whichever side believes turnout arithmetic favours it (TW-I-01).
The Tsai administration codified the phase-out — the January 2017 Electricity Act amendment wrote "all nuclear generation ceases by 2025" into law (Article 95-1) alongside a 20-per-cent-renewables / 50-per-cent-LNG / 30-per-cent-coal 2025 target mix [TBD-VERIFY: the 20/50/30 formulation] — and the referendum instrument promptly destabilised it. The November 2018 "Go Green with Nuclear" (以核養綠) referendum repealed Article 95-1 with roughly 59 per cent support; the government deleted the article but achieved the same phase-out through licence expiry, since 40-year operating licences lapsed one by one and Taipower filed no timely extensions. The December 2021 referendums then rejected restarting Lungmen and rejected relocating the Datan third LNG terminal away from the algal reef — a split verdict that ratified the LNG-heavy transition while burying the fourth plant. Energy policy was made and unmade by referendum four times in four years.
The implementation record of the energy transition is genuinely mixed: a real offshore-wind industry was built from zero, but renewables missed the 2025 target badly; LNG dependence deepened on schedule; Taipower absorbed catastrophic losses after the 2022 global price shock; and three major blackouts (15 August 2017, 13 May 2021, 3 March 2022) made grid fragility a permanent political fact. Renewables reached roughly 12 per cent of generation by 2024–2025 against the 20-per-cent target [TBD-VERIFY]; offshore wind installed approximately 3 GW against the 5.7 GW 2025 target [TBD-VERIFY], with localisation requirements inflating costs; ground-mounted solar collided with fishpond and farmland communities. Taipower's accumulated losses passed NT$400 billion [TBD-VERIFY] as the state absorbed imported-fuel inflation rather than pass it to households and fabs — an implicit industrial subsidy with mounting fiscal weight.
On 17 May 2025 Taiwan became nuclear-free — Maanshan unit 2 disconnected as the last operating reactor — at precisely the moment the political consensus was reversing. Four days earlier, on 13 May 2025, the KMT-TPP legislative majority (TW-D-02) amended the Nuclear Reactor Facilities Regulation Act to permit licence extensions of up to 20 years, including applications after expiry [TBD-VERIFY: amendment provisions]. The 23 August 2025 referendum on restarting Maanshan won approximately three-quarters of votes cast but failed the 25-per-cent-of-electorate validity threshold [TBD-VERIFY: tallies of roughly 4.3 million yes / 1.5 million no against a threshold near 5 million]. The Lai government repositioned: "no ideological opposition" to advanced nuclear provided safety, waste, and social consensus conditions are met [TBD-VERIFY: Premier Cho Jung-tai and President Lai statements] — a formula that abandons the DPP's founding plank in substance while preserving it in procedure.
The post-2022 strategic environment — Russia's weaponisation of energy in the Ukraine war, PLA blockade rehearsals, and US pressure on Taiwanese resilience [TBD-VERIFY: the US-pressure dimension rests on reported official statements rather than published demands] — recast nuclear power from environmental question to survival question, and both readings of the phase-out remain at full strength. The phase-out-was-right reading holds that reactors on faults near Taipei, with no high-level waste solution and a Tao-injustice unresolved, were a risk democratic Taiwan was correct to retire, and that blockade scenarios doom an island grid regardless of fuel mix. The phase-out-was-strategic-error reading holds that Taiwan voluntarily surrendered its only domestically stored baseload — months of fuel on-site versus 11 days of gas — during the highest-threat decade since 1958. The corpus records both at full strength; the August 2025 referendum showed a public majority for the second reading that could not clear the institutional bar.
Comparatively, Taiwan ran the German Energiewende playbook in the strategic position of 1948 West Berlin. Japan restarted reactors after Fukushima; South Korea's Moon-era phase-out was reversed by Yoon within five years; Germany completed its exit in April 2023 amid the Ukraine energy shock and a continental grid. Taiwan is the only democracy to have completed a nuclear phase-out while facing an explicit threat of blockade and invasion, with no neighbours to import electricity from — an island grid with no interconnectors. What distinguishes the Taiwanese case in the democratic-energy-transitions literature is the existential-security overlay: every energy choice is simultaneously a climate choice, a price choice, and a war-survivability choice, and the three logics do not point the same way.

2. The Energy Predicament: Structural Exposure and the Demand Side

2.1 An Island Grid with Nothing Underneath It

Taiwan has effectively no domestic fossil-fuel endowment. The small coal seams of the Japanese colonial era were exhausted by the 1980s; offshore gas around Hsinchu and Miaoli produces marginal volumes; there is no oil. The Energy Administration's standard statistic — approximately 97.5–98 per cent of primary energy supply imported [TBD-VERIFY: the precise annual figure; Energy Statistics Handbook editions 2015–2024 cite 97.3–98.0 per cent] — places Taiwan alongside Japan, South Korea, and Singapore among the most import-exposed significant economies. What distinguishes Taiwan from those comparators is not the percentage but the threat model. Japan and South Korea plan against price shocks and chokepoint disruptions; Taiwan must additionally plan against a state actor with declared intent to absorb it, a navy and air force rehearsing encirclement (TW-C-06), and a doctrine — visible in the 2022 and 2024 PLA exercise patterns — in which quarantine or blockade is among the most-discussed coercion options short of invasion (TW-F-04).

The grid compounds the exposure. Taiwan is an electrical island in the literal sense: no interconnectors, no neighbours to import from in an emergency, no continental balancing area to lean on as Germany leaned on France and Scandinavia through its own transition. Roughly 23 million people and the world's most strategically significant manufacturing cluster run on a single synchronous system whose north-south transmission corridor is itself a studied vulnerability — the 1999 Chi-Chi earthquake's collapse of a single transmission tower at Zhongliao blacked out much of the island, an event that grid planners still cite as the founding demonstration of N-1 fragility.

The fuel-stockpile arithmetic is the core of the security framing. Oil reserves, governed by the Petroleum Administration Act's stockpile mandates, run to roughly five months [TBD-VERIFY: commonly cited at 146 days combined government and operator stocks]. Coal stocks at power-plant yards run around 39 days [TBD-VERIFY]. Liquefied natural gas — the fuel the energy transition deliberately expanded to roughly half of generation — is the soft point: storage capacity equating to approximately 11 days of summer demand, with figures cited between 8 and 14 days depending on season and assumptions [TBD-VERIFY: the 11-day figure is the most-circulated, attributed to Economic Ministry statements circa 2022–2024]. Government targets to raise gas storage to 14 days of capacity by 2027 [TBD-VERIFY: target dates cited variously as 2027 and 2030] were adopted precisely because the blockade literature seized on the number. LNG cannot be stockpiled like coal or oil — it boils off, terminals are conspicuous targets, and LNG carriers are among the first vessels insurers withdraw from contested waters. The structural consequence: the more successfully Taiwan executed its gas-heavy energy transition, the shorter its survivable horizon under blockade became. This inversion — transition success as security regression — is the analytical heart of the post-2022 debate.

2.2 The Demand Side: The Fabs as the Grid's Centre of Gravity

Taiwanese electricity consumption is unusually industrial. Industry takes well over half of total consumption [TBD-VERIFY: Taipower statistics cite the industrial share around 55–57 per cent], and within industry the electronics-and-semiconductor complex is the dominant and fastest-growing bloc. TSMC alone consumed an estimated 7.5–9 per cent of all Taiwanese electricity by 2022–2023 [TBD-VERIFY: TSMC's ESG disclosures and third-party estimates; Greenpeace East Asia estimated ~7.5 per cent for 2022], a share projected to reach 12–15 per cent by around 2030 [TBD-VERIFY] as the N2-and-below nodes — extreme-ultraviolet lithography is voraciously power-hungry — and AI-era capacity expansions documented in TW-G-01 and TW-G-04 come online. Add UMC, Micron's Taiwan DRAM fabs, the panel makers, and the packaging-and-testing tier (TW-G-02), and the strategic sector's grid weight is larger still.

This produces three distinctive dynamics. First, demand growth returned. Taiwanese electricity demand was near-flat through the 2010s; the AI-investment surge from 2023 onward restored growth projections of roughly 2.5–2.8 per cent annually through 2033 [TBD-VERIFY: Energy Administration long-term load forecasts], with semiconductor expansion the explicit driver. An energy transition designed in 2016 against flat demand had to deliver against rising demand. Second, quality requirements tightened. Advanced fabs are intolerant of voltage sags measured in milliseconds; each of the major blackouts (Section 5.4) produced wafer-scrap claims and, more consequentially, board-level questions inside TSMC about Taiwan-site concentration — feeding the diversification logic of TW-G-03. Third, the corporate energy agenda became quasi-sovereign. TSMC's RE100 commitment (100 per cent renewables by 2040, accelerated from 2050) made it the off-taker of most of Taiwan's offshore-wind output — including the single largest corporate renewable power-purchase agreement in the world at signing, with Ørsted's Greater Changhua 2b/4 [TBD-VERIFY: 920 MW, signed 2020] — so that the renewables build-out is, in practice, substantially a TSMC supply-chain project.

The pairing with TW-N-01's silicon-shield framing is direct. The shield argument holds that the world's dependence on Taiwanese fabs deters attack and guarantees rescue. The energy predicament is the shield's underside: the fabs that constitute the shield are the grid's largest load, the grid runs on fuels with days-to-weeks of storage, and a blockader need not bomb a single fab to idle them all. Analysts began pairing the phrases — silicon shield, energy Achilles' heel — in the post-2022 literature, and the pairing now structures how allied capitals assess Taiwanese resilience (Section 7).

2.3 The Institutional Architecture: Taipower and the Administered Price

The institution at the centre of every chapter of this story is the Taiwan Power Company (台灣電力公司), the state-owned vertically integrated monopoly founded in 1946 on the assets of the colonial Taiwan Electric Power Company. Taipower built the nuclear fleet, owns the grid, operates most thermal capacity, and is the counterparty to the renewables programme. The January 2017 Electricity Act amendment liberalised green-energy generation and direct corporate sales but deliberately left Taipower's integrated structure intact — Taiwan chose energy transition without market liberalisation, a sequencing choice with large consequences. Because Taipower remains the buyer and seller of nearly everything, electricity pricing is a Cabinet-level political decision taken through an electricity-price review committee, and successive governments of both parties have suppressed tariffs below cost — for households as welfare policy, for industry as competitiveness policy. The arrangement functioned tolerably while fuel was cheap; after 2022 it converted global LNG inflation directly into Taipower balance-sheet destruction (Section 5.3) and made the utility's solvency a recurring legislative hostage in the budget battles documented in TW-D-06.

3. The Nuclear Build and Its Politics (1978–2000)

3.1 The Developmental-State Programme: Six Reactors at Three Plants

Nuclear power came to Taiwan as high politics of the authoritarian developmental state. The First Nuclear Power Plant at Chinshan (金山, Shimen district, on Taiwan's northern tip) was one of Chiang Ching-kuo's Ten Major Construction Projects (十大建設) announced in 1973 — the same infrastructure programme that produced the freeway, the steel mill, and the container port, and that frames the KMT's self-narrative of economic statecraft. Chinshan unit 1 entered commercial operation in December 1978, unit 2 in 1979; the Second Nuclear Power Plant at Kuosheng (國聖, Wanli, also on the northern coast) followed with units in 1981 and 1983; the Third at Maanshan (馬鞍山, Hengchun peninsula, the island's southern tip) in 1984 and 1985. All six reactors were US-supplied — General Electric boiling-water reactors in the north, Westinghouse pressurised-water reactors at Maanshan — procured in the era when US nuclear-export relationships also served as the channel through which Washington policed and ultimately terminated Taiwan's covert weapons-related research at the Institute of Nuclear Energy Research [TBD-VERIFY: the 1988 Chang Hsien-yi defection and the forced shutdown of the programme; documented in declassified US records but peripheral to this document's scope].

By the mid-1980s the six reactors supplied somewhere between two-fifths and half of Taiwanese electricity [TBD-VERIFY: peak nuclear share circa 1985 cited between 41 and 52 per cent in different series] — proportionally among the highest nuclear dependencies in the world at the time. The programme's politics were the politics of the era: sites selected by Taipower and the Atomic Energy Council without consultation, local objection legally impossible under martial law, and the entire apparatus — utility, regulator, ruling party — institutionally fused. The Atomic Energy Council was simultaneously promoter and regulator of nuclear power, a dual mandate that the anti-nuclear movement would later make a central target and that was not resolved until the council's 2023 reorganisation into the Nuclear Safety Commission [TBD-VERIFY: reorganisation date].

3.2 Orchid Island: The Founding Injustice

In 1982 Taipower began shipping low-level radioactive waste to a storage facility at the southern end of Orchid Island (Lanyu, 蘭嶼), the small island off Taiwan's southeast coast that is the homeland of the Tao (Yami) people, the most geographically and culturally distinct of Taiwan's indigenous groups. The facility had been sited there by a 1974–1978 decision process in which the Tao were never meaningfully consulted; the enduring account — that islanders were told the construction was a fish cannery — is repeated in virtually every treatment of the episode [TBD-VERIFY: the cannery account is canonical in movement historiography and was substantially credited by the 2018 truth-investigation report, but the precise documentary basis remains debated]. By 1996, when shipments stopped under protest blockades, more than 97,000 barrels had accumulated [TBD-VERIFY: barrel count commonly cited at ~97,000–100,000].

The Tao anti-nuclear protests — beginning with the February 1988 "Drive Out the Evil Spirits" (驅逐惡靈) demonstration, led by figures including the writer Syaman Rapongan and channelled through the island's churches — were among the first sustained indigenous political mobilisations of the democratising era, and they fused two currents that have remained joined ever since: anti-nuclear politics and indigenous-rights politics. Successive governments promised removal of the waste; none has achieved it, because no Taiwanese locality will accept a replacement site — the same siting paralysis that blocks any long-term solution for the high-level spent fuel accumulating in pools at the three main plants. The Tsai administration's 2016 apology to indigenous peoples named Lanyu explicitly; a 2018 investigation report acknowledged the consent failure; and in 2019 the government granted NT$2.55 billion in retroactive compensation [TBD-VERIFY: compensation figure and date], which Tao assemblies accepted in part and contested in part. For the anti-nuclear movement, Orchid Island settled the moral question permanently: Taiwanese nuclear power was born of the same authoritarian disregard as the political order the democracy movement overthrew.

3.3 Chernobyl, the DPP, and the Green in Green Politics

The Chernobyl disaster of April 1986 detonated in Taiwanese politics within months of the DPP's September 1986 founding and a year before martial law lifted in July 1987. The Taiwan Environmental Protection Union (台灣環境保護聯盟), founded in 1987 by academics including the physicist Chang Kuo-lung, made opposition to the planned fourth nuclear plant its signature cause; local self-help associations formed around the three operating plants and the proposed Gongliao site. The DPP wrote the nuclear-free-homeland aspiration into its party platform in the late 1980s [TBD-VERIFY: the anti-nuclear clause entered the DPP programme circa 1988; precise party-congress date], making it — alongside self-determination — one of the party's two constitutive commitments. The "green" in Taiwanese green politics is thus literal: the DPP's colour, its environmental founding plank, and its grassroots geography (the anti-plant self-help associations were early DPP local organising nodes) grew together. This is the essential background to everything after 2016: when the Tsai government legislated the phase-out, it was discharging a founding promise, and when the Lai government loosened it (Section 6), it was renegotiating party identity, not merely policy.

Through the 1990s the movement scored its first institutional victories as democratisation opened the system: the 1994 Gongliao local referendum (96 per cent against the fourth plant, legally non-binding, turnout 58 per cent [TBD-VERIFY]), repeated DPP-sponsored budget-deletion motions in the Legislative Yuan, and a 1996 legislative resolution to halt the project that President Lee Teng-hui's administration overrode. The KMT majority sustained the project; construction of the Fourth Nuclear Power Plant — Lungmen (龍門), at Gongliao on the northeast coast, two 1,350-MW GE-Hitachi advanced boiling-water reactors — finally began in March 1999.

3.4 The 2000 Halt-and-Restart Crisis: Energy Policy Meets Divided Government

Chen Shui-bian's March 2000 victory — the first non-KMT presidency — put a movement party with an abolitionist platform in charge of a half-built reactor complex, facing a KMT-dominated Legislative Yuan. On 27 October 2000, after a review led by Premier Chang Chun-hsiung (the economics minister's recommendation to continue having been overridden, and the announcement timed — gratuitously, in KMT eyes — minutes after Chen met KMT chairman Lien Chan), the Executive Yuan announced suspension of Lungmen construction [TBD-VERIFY: sequence and timing details]. The legislature erupted: recall motions against Chen were drafted, cross-party talks collapsed, and the KMT-PFP majority petitioned the Council of Grand Justices. Interpretation No. 520 (15 January 2001) held that a statutory budget already passed by the legislature could not be unilaterally suspended by the Executive Yuan without reporting to and negotiating with the Legislative Yuan — a landmark of the five-yuan separation-of-powers jurisprudence documented in TW-I-01. Lacking the votes and facing economic-confidence damage amid the dot-com downturn, the Chen government announced resumption of construction on 14 February 2001, paired with a face-saving legislative resolution endorsing the nuclear-free homeland as a long-term national aspiration — language later echoed in the 2002 Environmental Basic Law's Article 23 [TBD-VERIFY: article number].

The 2000–2001 crisis set the template for the next quarter-century. First, it established that energy policy under divided government is decided by institutional attrition, not deliberation — a lesson re-run in 2014, 2018, 2021, and 2025. Second, it poisoned Lungmen's engineering: the halt scattered contractors, the restart proceeded under a design-change regime in which Taipower self-modified a turnkey design, and the project's subsequent two decades of delays, budget expansions (to over NT$280 billion [TBD-VERIFY]), and safety-review controversies trace substantially to the 2000–2001 rupture. Third, it taught the DPP that the phase-out could not be won frontally against a KMT legislature — shaping the Tsai-era strategy of achieving it through licence expiry rather than confrontation. Throughout the Chen presidency (TW-A-03) the issue then went quiet: the six operating reactors ran on, Lungmen inched forward, and energy politics waited for an external shock.

4. Fukushima and the Phase-Out Decision (2011–2021)

4.1 The Fukushima Shock's Taiwan Effect

The 11 March 2011 Tōhoku earthquake, tsunami, and triple meltdown at Fukushima Daiichi struck the one democracy whose nuclear geography most resembles Taiwan's: an earthquake-and-tsunami coast, ageing GE boiling-water reactors of the same families as Chinshan and Kuosheng, and dense population nearby. The Taiwanese specifics were worse on the last point — Chinshan and Kuosheng sit roughly 20–30 kilometres from Taipei, and analyses circulating after 2011 (a widely cited Nature commentary ranked the two plants among the world's highest-risk for surrounding population [TBD-VERIFY: the 2011 Nature analysis identified Kuosheng and Chinshan among reactors with the largest 30-km populations]) made the evacuation question unanswerable: there is no plan under which greater Taipei's millions move.

Fukushima converted the anti-nuclear cause from a green-camp marker into a cross-cutting civic movement. The annual March anti-nuclear marches grew from tens of thousands in 2012 to a peak of over 200,000 across Taipei, Taichung, Kaohsiung, and Taitung on 9 March 2013 [TBD-VERIFY: organiser figures ~220,000; police figures lower] — at the time among the largest demonstrations in Taiwanese history. Celebrities, parents' groups, and business figures joined; polling showed majorities for halting Lungmen even among KMT identifiers [TBD-VERIFY]. The Ma government, committed to completing Lungmen but unwilling to own it outright, attempted a jiu-jitsu move in 2013: KMT caucus whip Lin Hung-chih proposed a national referendum on halting the plant, calculating that the Referendum Act's double threshold (50 per cent turnout, then majority) — the "birdcage" design that had defeated every national referendum 2004–2008 — would produce a failed referendum and thereby a mandate to proceed. The opposition denounced the gambit; the referendum was never held; the manoeuvre's main effect was to make Referendum Act reform itself a movement demand, with consequences in 2017 (Section 4.3).

4.2 April 2014: Lin Yi-hsiung and the Mothballing

The climax came weeks after the Sunflower occupation of the Legislative Yuan ended (TW-B-05). On 22 April 2014, Lin Yi-hsiung — the former DPP chairman whose mother and twin daughters were murdered in 1980 during his imprisonment after the Kaohsiung Incident, and who had spent two decades as the anti-nuclear movement's moral anchor, including leading the 1994–2003 "walks" for a Lungmen referendum — began an indefinite hunger strike at the Gikong Presbyterian church, the site of the 1980 murders, demanding a halt to Lungmen. The symbolic compression was overwhelming: the movement's most sacrificial figure, the democracy struggle's most haunting site, and a public already mobilised by the Sunflower spring. Tens of thousands occupied Zhongxiao West Road outside Taipei Main Station on 27–28 April; police cleared the crowd with water cannon overnight [TBD-VERIFY: clearance details].

On 27 April 2014 the Ma government capitulated in substance: Premier Jiang Yi-huah announced that Lungmen unit 1 would be sealed and mothballed (封存) after completing safety checks, unit 2's construction halted, and any future activation decided by national referendum [TBD-VERIFY: announcement formula]. Lin ended his strike on 30 April. Unit 1's mothballing was completed in July 2015, with fuel rods later shipped back to the United States from 2018 [TBD-VERIFY]. A KMT government had, under street pressure, terminated the KMT's flagship energy project — without resolving anything: the plant stood intact, the sunk cost unbooked, the referendum promise live ammunition for later (Section 4.4).

4.3 The Tsai Codification: Nuclear-Free Homeland 2025

Tsai Ing-wen campaigned in 2016 on completing the founding promise: a nuclear-free homeland by 2025, the year the last 40-year licence (Maanshan unit 2) would expire — a deadline chosen because it required no early closures, only the non-renewal of licences. The January 2017 Electricity Act amendment wrote it into statute as Article 95-1: "all nuclear-energy-based power-generating facilities shall wholly stop running by 2025." Around the deadline the administration built the Energy Transition (能源轉型) framework: a 2025 generation mix of 20 per cent renewables, 50 per cent LNG, and 30 per cent coal [TBD-VERIFY: the 20/50/30 targets, stated in the 2017 Guidelines on Energy Development and the 2018 Energy Transition White Paper]; a 5.7-GW (later 5.6-GW) offshore-wind target and 20-GW solar target for 2025 [TBD-VERIFY]; new LNG receiving terminals at Guantang (Taoyuan) and Taichung plus Keelung's Xiehe conversion; and the Electricity Act's green-energy liberalisation enabling corporate power-purchase agreements — the channel through which TSMC would become the renewables programme's anchor buyer (Section 2.2).

The design's political logic was expiry rather than confrontation: no reactor would be closed by decree; Taipower would simply not file licence-extension applications within the statutory five-years-before-expiry window, and the fleet would sunset on its own schedule — Chinshan units in December 2018 and July 2019, Kuosheng in December 2021 and March 2023, Maanshan in July 2024 and May 2025. The design's substantive gamble was that renewables and LNG infrastructure would arrive on schedule to replace roughly 10–12 per cent of generation [TBD-VERIFY: nuclear share circa 2016] while demand stayed flat. Neither assumption fully held (Section 5).

4.4 The Referendum Whiplash: 2018 and 2021

The Tsai administration's December 2017 Referendum Act amendment — a Sunflower-era democratisation commitment — lowered the proposal and signature thresholds and replaced the 50-per-cent-turnout requirement with a validity threshold of 25 per cent of the electorate voting yes. It thereby handed the energy debate to direct democracy, and the instrument promptly turned against its reformers.

In the 24 November 2018 referendums, held alongside the nine-in-one local elections in which the DPP was routed, the "Go Green with Nuclear" (以核養綠) initiative led by nuclear scientist Huang Shih-hsiu won approximately 5.9 million yes votes against 4.0 million no (around 59 per cent of valid votes) [TBD-VERIFY: tallies], abolishing Electricity Act Article 95-1. Two companion measures — reducing coal generation by 1 per cent annually and halting new coal plants — also passed, mathematically squeezing the system from both ends. The government's response defined the next phase: it deleted Article 95-1 as legally required, but declared the nuclear-free-homeland policy unchanged, and the licence-expiry mechanism rolled on because the statutory extension-application windows for Chinshan and Kuosheng had already lapsed [TBD-VERIFY: the Atomic Energy Council's position that applications were time-barred]. Phase-out by deadline became phase-out by administrative default — legally robust, democratically awkward, and the precise grievance the KMT-TPP majority would target in 2025.

The 18 December 2021 referendums ran the experiment in reverse. Four questions, two on energy: restart Lungmen unit 1 (proposed by Huang Shih-hsiu), and relocate the third LNG receiving terminal away from the Datan algal reef at Taoyuan (proposed by reef-conservation campaigner Pan Chong-cheng, backed by the KMT). The algal-reef question split the environmental movement itself — a 7,000-year-old biogenic reef hosting endangered species against the LNG terminal needed to retire coal at Taichung and replace nuclear; the government compromised mid-campaign by moving the terminal design further offshore [TBD-VERIFY: the "outward relocation" plan, extending the project by 2.5 years]. On the day, with turnout around 41 per cent, all four referendums failed: Lungmen restart drew approximately 3.8 million yes against 4.26 million no, and the reef relocation similarly fell short [TBD-VERIFY: tallies]. The Tsai government read the result as ratification of the entire energy-transition architecture; the opposition read it as a turnout artefact. Both readings fed the conclusion documented in TW-I-01: the post-2017 referendum instrument had made energy policy the most plebiscite-whipsawed domain in Taiwanese governance — codified 2017, decodified 2018, re-ratified 2021, re-contested 2025 — with each verdict binding for two years under the Referendum Act and revisable thereafter.

5. The Implementation Record (2016–2026)

5.1 The Renewables Build-Out: A Real Industry, A Missed Target

The offshore-wind programme is the energy transition's genuine institutional achievement and its clearest cost lesson. From a standing start — Formosa 1, Taiwan's first commercial offshore wind farm (128 MW, Swancor/Ørsted/JERA), entered operation in late 2019 — the Ministry of Economic Affairs ran a three-stage architecture: demonstration projects, a 2018 zonal-development allocation of approximately 5.5 GW across selection (with feed-in tariffs among the world's most generous, around NT$5.8/kWh initially [TBD-VERIFY]) and auction tracks (with winning bids stunningly low, near NT$2.2–2.5/kWh [TBD-VERIFY]), and post-2021 Round 3 block auctions. The majors came: Ørsted's Greater Changhua complex, wpd/Skyborn's Yunlin, Copenhagen Infrastructure Partners' Changfang-Xidao and Zhong Neng, Northland Power's Hai Long. The Taiwan Strait's wind resource is world-class; the strait is also, as every developer's risk committee noted, a potential war zone — a fact that surfaced in financing premia and insurance terms [TBD-VERIFY: anecdotal in industry reporting].

The localisation requirements — mandatory Taiwanese content in towers, foundations, cables, and eventually nacelle assembly, designed to seed a domestic supply chain around the Taichung port cluster and Century Wind Power, Swancor, and CSBC — became the programme's central controversy. Developers attributed multi-year delays and substantial cost inflation to immature local suppliers; the government defended localisation as industrial policy in the Hsinchu tradition (TW-G-04); the Yunlin project's near-collapse and repeated rescue restructurings [TBD-VERIFY] became the cautionary case. Under EU trade pressure and developer attrition, Round 3 rules progressively relaxed localisation [TBD-VERIFY: the 2024 industrial-relevance scheme revisions following EU WTO consultations]. Installed offshore capacity reached roughly 2.25 GW by end-2023 and around 3 GW during 2024–2025 [TBD-VERIFY], against the 5.7-GW 2025 target — a real fleet, materially late.

Solar tells a harder land story. The 20-GW 2025 target (3 GW rooftop, 17 GW ground-mounted, later rebalanced toward rooftop) collided with the fact that Taiwan has almost no spare flat land: ground-mounted solar went to fishponds, marginal farmland, and salt flats, generating sustained conflict over the "fishery-electricity symbiosis" (漁電共生) model in Tainan and Chiayi — aquaculture communities contesting displacement, ecologists contesting wetland-bird habitat loss, and corruption cases around land-conversion permits [TBD-VERIFY: prosecutions including the 2022–2023 Tainan solar-permit cases]. Installed solar reached approximately 12–14 GW by 2024–2025 [TBD-VERIFY], again materially short. Total renewables generation share reached roughly 12 per cent in 2024–2025 [TBD-VERIFY: 2024 figures cited around 11.6–12.5 per cent] against the 20-per-cent target, with the government re-dating the 20-per-cent milestone to 2026–2027 [TBD-VERIFY]. The honest summary: the transition built Asia's second-largest offshore-wind fleet after China's, created a domestic industry, and missed its headline target by roughly eight percentage points — a gap filled by fossil fuels.

5.2 The LNG Deepening and the Terminal Politics

Gas-fired generation rose from roughly a third of output in 2016 to around 42–47 per cent by 2024–2025 [TBD-VERIFY], on trajectory toward the 50-per-cent design point. The physical bottleneck was regasification: two legacy terminals (Yung-An/Kaohsiung, 1990; Taichung, 2009) ran above nameplate for years [TBD-VERIFY], and the build-out — the contested third terminal at Guantang/Datan (operational 2025 [TBD-VERIFY]), the fourth at Keelung Xiehe (the subject of a 2023–2024 local-referendum fight of its own [TBD-VERIFY]), the fifth at Taichung — ran through exactly the coastal-siting conflicts (algal reef, fisheries, port communities) that nuclear siting once ran through. The pattern deserves the comparative note it gets in the literature: the energy transition did not end facility-siting conflict in a dense island democracy; it relocated it from reactors to terminals, substituting climate-aligned infrastructure with war-vulnerable infrastructure.

5.3 The Price Shock and Taipower's Balance Sheet

Russia's February 2022 invasion of Ukraine sent Asian spot LNG above US$50/MMBtu and re-priced term contracts across the board. Because Taiwanese electricity tariffs are administered (Section 2.3), the government chose — across both the Tsai and Lai cabinets — to absorb rather than fully pass through: tariff rises in 2022–2024 were repeated but cumulatively far below cost recovery, weighted onto large industrial users (the fabs took double-digit increases [TBD-VERIFY: the October 2022 and April 2024 industrial-tariff adjustments]) while household rates were shielded. Taipower's accumulated losses exceeded NT$400 billion by 2024 [TBD-VERIFY: figures cited NT$382–430 billion across 2023–2025 reporting], met by repeated capital injections and subsidy appropriations — including injections the KMT-TPP legislative majority cut or froze during the 2024–2026 budget confrontations (TW-D-06), turning the utility's solvency into a hostage of divided government. The strategic irony was widely noted: a state that subsidises electricity below cost for the world's most profitable chipmaker is running an implicit industrial subsidy through a bankrupt utility, while the cheap power dampens the price signal that would otherwise drive industrial efficiency and demand response.

5.4 Blackouts and the N-1 Question

Three events fixed grid fragility in public memory. On 15 August 2017, a contractor error at the Datan gas plant — a mistakenly closed gas-supply valve during maintenance — tripped six units and cut power to 6.68 million households (the "815 blackout"); Economics Minister Lee Chih-kung resigned that evening [TBD-VERIFY: details]. On 13 May 2021 a grid mishap at the Hsinta plant (Kaohsiung) forced rotating outages to roughly 4 million households, repeated in a second event on 17 May 2021 [TBD-VERIFY]. On 3 March 2022 a switching error at Hsinta cascaded into the "303 blackout" affecting around 5.5 million households, including the Kaohsiung industrial belt; Economics Minister Wang Mei-hua offered her resignation, which Premier Su Tseng-chang declined [TBD-VERIFY]. Each event was proximately operational, not a generation shortfall; each was nonetheless politically metabolised as evidence on the phase-out — the opposition reading them as a thinning system's loss of margin (reserve margins had at times dropped below 3 per cent in the late 2010s [TBD-VERIFY]), the government reading them as grid-architecture failures answerable by the post-2022 NT$564.5-billion grid-resilience programme [TBD-VERIFY: Taipower's ten-year grid resilience plan announced September 2022], which decentralises the trunk system precisely because both the blackout record and the blockade literature (Section 7.1) identified the centralised grid as the binding fragility. The deeper N-1 debate — whether an island system running single-digit reserve margins with no interconnectors can responsibly retire firm capacity on schedule — remained unresolved because the two camps could not agree on what the blackouts were evidence of.

5.5 17 May 2025: Nuclear-Free Taiwan, Achieved and Immediately Contested

The fleet sunset on schedule: Chinshan 1 (December 2018), Chinshan 2 (July 2019), Kuosheng 1 (December 2021, six months early on spent-fuel-pool saturation [TBD-VERIFY]), Kuosheng 2 (March 2023), Maanshan 1 (27 July 2024), and finally Maanshan 2, disconnected on 17 May 2025 [TBD-VERIFY: shutdown date]. Nuclear's generation share had fallen from roughly 12 per cent in 2016 to about 3 per cent in 2024 to zero. The nuclear-free homeland — the DPP's founding plank, four decades in the making — was achieved under a DPP president. It was achieved four days after the Legislative Yuan voted to make its reversal legally possible, and three months before a referendum in which a large majority of those voting asked for it back. Section 6 takes up that inversion.

6. The Reversal Debate (2024–2026)

6.1 The KMT-TPP Legislative Push and the May 2025 Amendment

The 2024 elections produced the hung Legislative Yuan documented in TW-D-02: a DPP presidency facing a KMT-TPP working majority that made energy one of its principal confrontation fronts, alongside the budget (TW-D-06) and the legislative-powers package. The TPP under Ko Wen-je and then Huang Kuo-chang had positioned itself as the party of technocratic nuclear pragmatism; the KMT had completed its journey from the party that mothballed Lungmen under protest to the party of unambiguous nuclear revival. On 13 May 2025 the bloc passed an amendment to the Nuclear Reactor Facilities Regulation Act (核子反應器設施管制法) permitting operating-licence renewals of up to 20 years per application and — the operative innovation — allowing applications to be filed even after a licence has expired or a reactor has shut down, dissolving the time-bar through which the phase-out-by-default mechanism had run [TBD-VERIFY: amendment provisions and vote count]. The Lai cabinet did not seek a constitutional-court challenge; Premier Cho Jung-tai stated the government would respect the law while insisting any restart must clear safety review by the Nuclear Safety Commission, waste-management planning, and "social consensus" [TBD-VERIFY: the three-conditions formula]. Four days later Maanshan 2 shut on schedule: the law now permitted what the calendar had just foreclosed.

6.2 The 23 August 2025 Referendum: A Majority That Did Not Count

The TPP caucus moved immediately to the plebiscite channel, and on 23 August 2025 Taiwan voted on the question, in substance: "Do you agree that the Third Nuclear Power Plant (Maanshan) should continue operating upon confirmation by the competent authority that there are no safety concerns?" The result compressed the entire post-2017 referendum pathology into one event: approximately 4.34 million yes votes against roughly 1.51 million no — about 74 per cent of votes cast — but turnout near 30 per cent left the yes total short of the validity threshold of 25 per cent of the electorate (approximately 5 million votes), so the referendum formally failed [TBD-VERIFY: official CEC tallies, turnout, and threshold figures]. A three-to-one majority of participating voters endorsed restart; the institution recorded a null. The DPP had quietly counselled abstention rather than opposition — the turnout-suppression strategy the birdcage-era KMT had pioneered — and the result therefore settled nothing while revealing much: polled and voting majorities now favoured nuclear continuation [TBD-VERIFY: 2024–2025 polling consistently showed majorities for retaining or restarting nuclear plants], and the nuclear-free homeland no longer commanded the public it was built for.

President Lai's same-evening response marked the doctrinal shift: the government "does not rule out advanced nuclear energy" provided nuclear safety, a waste solution, and social consensus are achieved, and Taipower would proceed with safety assessments of the shut Maanshan units [TBD-VERIFY: Lai's 23 August 2025 statement; Taipower subsequently announced a multi-year restart-feasibility assessment]. "No ideological opposition" — the phrase circulating from cabinet figures through 2024–2025 [TBD-VERIFY: attribution; variants appear in Cho Jung-tai's and economics-ministry statements] — became the formula by which a DPP government vacated the founding plank in substance while retaining procedural control over whether and when any reactor returns. As of mid-2026 no restart application had been approved; Taipower's Maanshan assessment, the seismic re-evaluation (a fault re-survey near the plant complicated the case [TBD-VERIFY: the Hengchun fault question]), and the spent-fuel storage bottleneck — dry-cask facilities at all three plants remained blocked by local-government permitting, principally New Taipei's decade-long obstruction [TBD-VERIFY] — meant any physical restart lay years out. The reversal, like the phase-out, proceeds by attrition.

6.3 The Arguments Recast: AI Demand, Net Zero, and the SMR Horizon

What changed between 2014 and 2025 was less the safety facts than the argument structure around them. Three new frames did the work. First, AI-era demand: the projection that semiconductor and data-centre growth would add the equivalent of several gigawatts of firm demand by 2030 (Section 2.2), endorsed by figures with green-camp credibility — most visibly Pegatron's Tung Tzu-hsien, a member of Lai's National Climate Change Committee, and former premier and Sunflower-era favourite figures arguing that renewables-plus-gas could not carry both decarbonisation and the fabs [TBD-VERIFY: Tung's and related committee statements; NVIDIA's Jensen Huang also publicly favoured Taiwanese nuclear supply]. Second, net zero: Taiwan's statutory 2050 net-zero commitment (the 2023 Climate Change Response Act) made the 80-plus-per-cent fossil grid of 2025 [TBD-VERIFY: fossil share ~83–84 per cent] an international-credibility problem — Taiwanese exporters face EU carbon-border accounting, and nuclear's reclassification as green in the EU taxonomy and at COP28 gave the revival case a climate vocabulary the 2014 movement never had to answer. Third, the SMR horizon: small modular reactors entered the debate as the form in which "advanced nuclear" could be endorsed without re-fighting the Chinshan-Kuosheng siting battles, with the National Climate Change Committee and Taipower commissioning SMR-feasibility studies [TBD-VERIFY] — though no SMR design was licensable in Taiwan as of 2026, making the SMR position, critics noted, a way of being pro-nuclear in the future tense.

6.4 Interim: The Central Election Commission and the Fourth Nuclear Referendum (June–August 2026)

The reversal debate acquired its second plebiscite within a year of the failed August 2025 vote. Through July and August 2026, opposition-aligned civic groups fast-tracked a fresh referendum proposal seeking to abolish the "Nuclear-Free Homeland" (非核家園) policy outright — distinct from the 2025 question's narrower Maanshan-restart framing — with the Legislative Yuan approving three opposition-backed referendum proposals for onward CEC review on 14 August 2026, alongside two non-energy questions (on judicial caning for certain offences and on traffic-fine reform) [TBD-VERIFY: precise sponsoring civic groups and the 14 July 2026 fast-track procedural sequence, per Focus Taiwan reporting]. On 28–29 August 2026 the Central Election Commission narrowed the field, approving the nuclear-policy referendum for the ballot while rejecting the caning and traffic-fine proposals as failing procedural or content review — six commissioners voting in favour with one abstention [TBD-VERIFY: exact CEC vote breakdown and rejection grounds]. The approved question asks, in substance, whether the government should abolish the Nuclear-Free Homeland policy to secure public health, stable electricity supply, affordable rates, and strengthened national-defence resilience [TBD-VERIFY: partial question text search-corroborated via Taipei Times reporting; full CEC-certified wording not yet confirmed against the primary gazette text]. It will appear on the 28 November 2026 ballot alongside the Nine-in-One local elections (TW-D-05) — Taiwan's fourth nuclear referendum since 2018, following the November 2018 "Go Green with Nuclear" vote, the December 2021 Lungmen-restart vote, and the August 2025 Maanshan-continuation vote, none of which has produced an institutionally binding restart.

The June–August 2026 window supplied two further data points sharpening the security and demand-side arguments documented in Sections 6.3–6.4. First, the Ministry of Economic Affairs's energy-demand projections — cited in the trade-press assessment of Taiwan's nuclear reversal published in August 2026 — put electricity-demand growth at approximately 2.5 per cent annually between 2026 and 2035, more than double the rate of the preceding decade, reinforcing the AI-and-fab-driven load-growth case for the pro-nuclear coalition [TBD-VERIFY: MOEA projection as reported by AmCham Taiwan's Taiwan Business TOPICS, August 2026; primary MOEA release not yet located]. Second, domestic commentary through early August 2026 continued to frame the restart question as institutionally unresolved even as public and legislative sentiment shifted: no Maanshan restart application had been filed, and the seismic and spent-fuel-storage bottlenecks identified in Section 6.2 remained the binding physical constraints regardless of the referendum outcome. The pairing of a fourth plebiscite with the Nine-in-One elections raises the same turnout-arithmetic question that has governed every energy referendum since 2018 (Section 4.4): a KMT-TPP-mobilised local-election turnout could plausibly clear the 25-per-cent-of-electorate validity threshold that defeated the August 2025 vote, in which case the November 2026 result — for the first time — could bind the government procedurally rather than merely register a "majority that did not count."

6.5 The Security Framing's Ascendancy

The decisive shift, however, was the security frame. After February 2022 — Russia's energy weaponisation, Europe's scramble, and the PLA's post-Pelosi normalisation of blockade-rehearsal exercises — energy stockpiles migrated into defence policy. The blockade arithmetic of Section 2.1 (11 days of gas [TBD-VERIFY]) became a standard talking point in legislative interpellations, US think-tank testimony, and the defence-resilience components of the special-budget debates (TW-D-06); the Lai administration's whole-of-society resilience committee included energy continuity among its workstreams [TBD-VERIFY]; and nuclear fuel's on-site endurance — a loaded core runs roughly 18 months, against days of gas and weeks of coal — became the revival case's sharpest single argument. The US dimension operated mostly as atmosphere rather than demand: reported remarks by US officials and former officials questioning the wisdom of the phase-out, American Institute in Taiwan engagement on energy resilience, and the Trump-2 administration's general pressure on allied energy security and interest in US nuclear exports [TBD-VERIFY: this dimension rests on reporting of informal statements; no published US demand to restart nuclear plants is documented]. The corpus should record the asymmetry: the security case for nuclear is now the strongest it has ever been in Taiwanese debate, and it is also true — as phase-out defenders reply — that six reactors and their spent-fuel pools on the coasts of a potential war zone are themselves targets, that Ukraine's Zaporizhzhia experience cuts both ways, and that no fuel mix makes an interconnector-less island grid survivable under sustained blockade. Both the phase-out-was-right and the phase-out-was-strategic-error readings are held by serious people with the same facts; the disagreement is over which risks a democracy facing invasion should choose to carry.

7. Energy as Cross-Strait and Comparative Question

7.1 The Blockade Mathematics

The post-2022 war-gaming and blockade literature converged on energy as Taiwan's decisive non-military vulnerability. The CSIS January 2023 invasion wargame (The First Battle of the Next War) modelled kinetic outcomes, but the companion analytical wave — CSIS, RAND, and Council on Foreign Relations blockade and quarantine studies through 2023–2025 [TBD-VERIFY: specific studies; CSIS published quarantine and blockade analyses in 2024] — concluded that a PRC energy quarantine (inspecting or turning back tankers and LNG carriers, possibly under a customs-enforcement pretext) could coerce Taiwan over weeks without firing on it, because LNG carriers' insurers and charterers withdraw at the first credible threat. Modelling exercises put electricity rationing within two to four weeks of a tight blockade, with gas-fired generation — half the system — failing first [TBD-VERIFY: scenario timelines vary widely]. Taiwanese counter-programming followed: raised stockpile mandates (14-day gas target [TBD-VERIFY]), the grid-resilience decentralisation programme (Section 5.4), hardening and dispersal studies for terminals, demand-rationing protocols within the whole-of-society resilience framework, and the argument — made inside the defence-spending debate documented in TW-D-06 — that a dollar of fuel storage may buy more deterrence-by-resilience than a dollar of munitions, since the PRC's cheapest coercion path runs through the energy system. The silicon-shield pairing completes the strategic picture (TW-N-01, TW-G-01): the fabs give the world a stake in Taiwan's lights staying on, and give Beijing a reason to prefer the coercion mode — blockade — that attacks the lights rather than the island.

7.2 The Comparative Frame: Three Democracies and a Pendulum

Taiwan's trajectory is best read against three comparators. Japan suffered the accident itself, shut its entire fleet, and then restarted: by 2025 fourteen reactors had returned under post-Fukushima regulatory standards [TBD-VERIFY: restart count], with the 2023–2025 GX (green transformation) policy extending licences and endorsing new builds — a phase-out reversed by energy-security and climate arithmetic without a single referendum, through bureaucratic-judicial channels that Taiwanese pro-nuclear advocates cite enviously and Taiwanese anti-nuclear advocates cite as exactly the democratic deficit Taiwan refused. South Korea ran the full pendulum on presidential alternation: Moon Jae-in's 2017 phase-out (cancelling new units, dramatised by the 2017 Shin-Kori 5/6 deliberative-polling reversal) was reversed wholesale by Yoon Suk-yeol from 2022, with nuclear exports made a national champion sector — demonstrating, as the Taiwanese debate noted, that phase-outs decided by single-party governments lack ratchet stability. Germany is the closest procedural cousin: a phase-out rooted in a decades-old green movement, codified post-Fukushima, and completed in April 2023 in the teeth of the Ukraine energy shock — but Germany ended its exit inside the European grid, with French nuclear and Norwegian hydro as neighbours, and even so paid a price-and-emissions cost that the Taiwanese revival camp treats as the controlled experiment proving the policy's expense.

The democratic-energy-transitions literature draws from these cases a general proposition: nuclear phase-outs are movement achievements that require decades of mobilisation to win and stable abundance to keep. Taiwan tests the proposition at its extreme. No other phase-out state faces an adversary that rehearses blockading it; no other operates an island grid with no interconnectors; no other hosts an industrial cluster whose continuity the entire world treats as systemically critical; and none made and unmade the policy through four national referendums in seven years. The Taiwanese case therefore does not show that democratic energy transitions fail — the offshore-wind fleet, the grid programme, and the completed phase-out are all democratic deliveries — but it shows that the existential-security overlay re-prices every element of the transition after the fact, and that direct-democracy instruments transmit that re-pricing into policy with a velocity party systems would otherwise buffer. The phase-out was a democratic achievement; so was the August 2025 majority asking for its reversal; Taiwanese institutions are now required to honour both.

8. Conclusion and Forward View

Across the 1978–2026 arc, Taiwanese energy policy has been the domain where every layer of the island's governance is exposed at once: the developmental state's capacity and its consent deficit (the six reactors, Orchid Island); the democracy movement's moral sources (Lin Yi-hsiung, the nuclear-free plank); divided government's attritional style (the 2000 Lungmen halt, Interpretation No. 520, the 2025 amendment); the referendum instrument's whipsaw (2018, 2021, 2025); the state-enterprise model's fiscal exposure (Taipower); and, above all, the cross-Strait threat that re-prices every domestic choice. The record as of mid-2026 admits a precise statement: Taiwan completed the democratic world's most contested nuclear phase-out on schedule, built a genuine renewables industry that missed its targets, deepened a gas dependence that the blockade literature identifies as its softest strategic point, and arrived at nuclear-free status at the moment its public, its legislature, and its strategic environment had turned against the policy — without any institutional verdict yet authorising its reversal.

The forward variables are identifiable. First, the Maanshan track: Taipower's safety and seismic assessments, the Nuclear Safety Commission's posture, the New Taipei and Pingtung dry-cask permitting blockages, and whether any licence-renewal application is actually filed and approved — the first physical test of the May 2025 amendment [TBD-VERIFY: assessment timelines run into 2027–2028 in Taipower statements]. Second, the November 2026 local elections and the 2028 presidential cycle, which will determine whether the "no ideological opposition" formula hardens into a restart programme (under any party) or remains a holding position. Third, the renewables and storage trajectory against AI-era demand — whether the re-dated 20-per-cent milestone, Round 3 offshore wind, geothermal and battery build-outs close the gap or whether fossil share remains above 80 per cent into the late 2020s. Fourth, the stockpile-and-resilience programme: the 14-day gas target, grid decentralisation, and the integration of energy continuity into defence planning, which will be tested rhetorically in every PLA exercise cycle and materially only in the contingency the entire architecture exists to deter. Fifth, the waste question, which no faction has solved and which constrains every nuclear future: until some Taiwanese jurisdiction accepts a final or interim high-level repository, both the phase-out's completion and any revival run through spent-fuel pools that are themselves a vulnerability.

For the corpus, TW-G-05 anchors the energy dimension of the Block-G economic architecture: TW-G-01 and TW-G-02 document the demand side whose growth drives the system; TW-G-03 documents the diversification that energy constraint partly motivates; TW-G-04 documents the industrial-policy machine that treats electricity as an input it has historically commanded. The security reading runs to TW-F-04 and TW-N-01 (energy as the blockade vulnerability beneath the silicon shield), the institutional reading to TW-I-01 (Interpretation No. 520 and the Referendum Act), and the political reading to TW-D-02 and TW-D-06 (the hung legislature and budget battles through which the reversal is being fought). The document's central analytical claim is the one the comparative section earns: Taiwan is the case where a democratic energy transition met an existential-security overlay, and the meeting resolved neither in favour of the other — which is why, uniquely among the democracies that debated nuclear power after Fukushima, Taiwan's debate is not over.

Document status: [DRAFT]. TBD-VERIFY tags mark capacity figures, referendum tallies, statutory provisions, dates, and attributed statements pending primary-source verification; the August 2025 referendum figures, the May 2025 amendment provisions, the Taipower financial figures, and the full CEC-certified text of the November 2026 nuclear referendum question are the highest-priority verification targets for the next fact-check wave. Version 2026-08-29.