The Energy Paradox of Taiwan’s Sovereign AI Ambition

With roughly half of Taiwan’s electricity generation reliant on imported liquefied natural gas (LNG), one-third of which is sourced from Qatar, the conflict in Iran and closure of the Strait of Hormuz serve as a stark reminder that Taiwan’s digital ambitions ultimately rest on fragile global energy supply chains. The stakes are rising as Taiwan accelerates its push into artificial intelligence. Last December, President Lai Ching-te’s announcement of Taiwan’s largest sovereign AI data center was framed as a milestone in Taiwan’s sovereign AI strategy by leveraging its global leadership in semiconductor manufacturing. Yet beneath the celebratory narrative lies an unresolved question: how does Taiwan plan to secure sufficient, affordable, and clean energy to power this vision of digital sovereignty?

Currently, Taiwan has 36 data centers with total capacity of about 60 MW. The government forecasts that the future capacity generated from AI applications and data centers will increase 6.5 times to 450 MW over the next three years. Yet to date, the public cannot access transparent assessments of how much of the predicted increase in baseload energy consumption due to AI hype is real and how that will affect household and industrial electricity consumption. The Ministry of Economic Affairs revised its 2025 estimate of overall demand growth for the next five years to 1.7 percent, down from its 2024 estimate of 2.5 percent, despite continued AI expansion. Moreover, projections for AI-specific electricity demand disappeared from the Ministry’s report entirely—leaving experts, utilities, and the public without a credible basis for long-term energy consumption forecasting.

The expected surge in energy demand for AI data centers is intensifying geographical imbalances in power supply and demand in Taiwan. Data centers and high-tech firms are concentrated in the north, while much of the island’s generation capacity is located in the south. Taiwan’s grid was not designed to continuously transmit large volumes of power over long distances, increasing risks of congestion, losses, and instability as energy demand grows in the north. To manage these constraints, the state-owned utility Taiwan Power Company (Taipower) has restricted new applications for power usage by data centers north of Taoyuan with capacity of more than 5 MW, unless new power capacity is built. This policy underscores not only the grid’s regional bottlenecks but also the physical limits of Taiwan’s electricity system; Taipower recently warned that this could lead to politically fraught competition for power between large technology firms and ordinary households.

The characteristics of AI power consumption risk growing the “baseload gap.” Unlike in traditional industry, in which power loads can fluctuate or be curtailed periodically, high-performance AI clusters operate continuously, requiring uninterrupted electricity for both computation and cooling. Many advanced economies have turned to nuclear power to supply affordable, stable, and low-carbon electricity to AI data centers. In Taiwan, however, nuclear power is a point of contention as questions about seismic risk and waste management remain unanswered. For many older Taiwanese, nuclear power also has a politicized legacy, shaped by nuclear development in the authoritarian era under the Chinese Nationalist Party (KMT) and the Democratic Progressive Party (DPP)’s founding platform of a nuclear-free Taiwan. These historical and ideological factors led to the “Nuclear-Free Homeland” policy launched in 2002, resulting in the retirement of Taiwan’s last reactor at the Maanshan Nuclear Power Plant in May 2025, even as electricity demand has continued to climb.

The tension between Taiwan’s energy realities and antinuclear policy came to a head last year, when over 74 percent of voters in an August 2025 referendum supported extending the life of the Maanshan plant. While the referendum ultimately failed due to insufficient voter turnout, the political signal was unmistakable. Even prior to the referendum, roughly two-thirds of the public supported reinstating nuclear power. In response, the Ministry of Economic Affairs softened its opposition to nuclear energy and agreed to review Taipower’s proposal to restart the Maanshan reactor, as President Lai recently announced in public for the first time. Even so, opinions among DPP supporters are divided, and progress toward reintroducing nuclear power remains slow; the government estimates that the Maanshan reactor would not restart until 2028 at the earliest. In Taiwan’s polarized political environment, contemporary partisan conflict and historical narratives continue to overshadow long-term energy planning.

To account for losing domestic power generation by going nuclear free, Taiwan has focused on scaling up renewable energy. Solar and wind generation continue to expand, but progress has been hampered by corruption scandals and repeated project delays, pushing the official 2025 target of 20 percent renewables in the energy mix beyond November 2026. More fundamentally, solar and wind power remain inherently intermittent. Until Taiwan has mature, large-scale, long-duration energy storage solutions, renewables cannot reliably power an Nvidia H200 or Blackwell-class cluster through the night. Because nuclear power remains contentious and unable to scale in the near term and renewable output remains variable, natural gas—almost entirely imported—has effectively become the island’s default baseload.

For a sovereign AI strategy, however, this deepening reliance on imported hydrocarbons constitutes a strategic vulnerability. Taiwan has sought to diversify its LNG import sources, primarily by purchasing more US gas. While this strategy may improve short-term energy resilience, deeper dependence on imported gas risks locking Taiwan into a carbon-intensive and geopolitically exposed energy trajectory. Taiwan’s LNG stockpile and storage capacity remain precariously limited, with the government aiming to build a gas reserve of only 14 days by 2027—an increasingly uncomfortable margin in light of the current de facto closure of the Strait of Hormuz and recurring PRC military drills in the Taiwan Strait. This creates a fundamental paradox: a “sovereign AI data center” powered by fuel vulnerable to maritime disruption is sovereign in name only.

In addition, the fiscal foundations of Taiwan’s energy system have become increasingly fragile. Since the start of the Russia–Ukraine war, the volatility of global LNG prices has sharply increased operating costs for Taipower, while a protracted conflict in Iran would further strain the global energy market. The government has sought to shield households from these price shocks through subsidies to Taipower and freeze of electricity rate, but this approach has reached its political and fiscal limits. In early 2025, the opposition-controlled Legislative Yuan blocked a proposed NT$100 billion (US$3.1 billion) subsidy intended to offset Taipower’s losses, which accumulated to more than NT$350 billion (US$11.0 billion) at the end of January this year.

As a result, electricity pricing and fuel diversification have now become politically radioactive. Public tolerance for increasing electricity prices is low, especially as increases are perceived to serve the highly profitable AI industry rather than ordinary households. In Taiwan, electricity pricing has long been treated as a social contract rather than a market signal, preventing the government from allowing prices to reflect the true cost of new power generation, grid upgrades, and energy security investments. This disconnect complicates the financing and justification for building next-generation energy infrastructure, creating paralysis in long-term capacity planning at a moment when the AI sector requires predictability, scale, and reliability.

The gap is widening between Taiwan’s AI ambitions and its energy preparedness. Policymakers must treat industrial policy and energy system planning not as separate domains but as two sides of the same coin. Recent steps, such as tightened approval requirements with stricter energy efficiency oversight for data-center projects, proposed behind-the-meter storage subsidies for AI facilities, and Taipower’s push for a “Power Couple” colocation strategy to pair AI data centers with newly built renewable and battery solutions, are welcome moves. Yet they remain incremental relative to the scale of the challenge. Comprehensive policy planning, stronger buy-in from industry on energy governance, and transparent communication with local communities are critical to managing the socioeconomic impacts of the buildup of AI infrastructure and its associated energy demand. Without a credible strategy to align AI expansion with energy resilience, Taiwan risks allowing its sovereign AI vision to outpace the systems required to sustain it, leaving Taiwan with the ability to manufacture the world’s most advanced chips but insufficient power to run them.

This dilemma is not unique to Taiwan; governments across Asia are pursuing AI to drive economic upgrading and leverage strategic advantages but struggle to internalize the energy implications of data-intensive infrastructure. In the AI era, leadership will not belong to those with the most advanced chips alone but to those who can power them securely, affordably, and sustainably. In a period defined by geopolitical competition and energy uncertainty, energy resilience is not an afterthought—it will be the bedrock of digital sovereignty and technological power for the coming decades.

The author thanks the anonymous reviewers for their comments on a draft of this piece. 

Author

Siwei Huang

Director of Engagement, Center for Asia-Pacific Resilience and Innovation

Biography

Siwei Huang is director of engagement of the Center for Asia-Pacific Resilience and Innovation (CAPRI), where he leads engagement initiatives with diverse stakeholders worldwide and conducts interdisciplinary research on public policy and international affairs in the Asia-Pacific region. He is leading CAPRI’s multiyear regional research project on pathways for Asia-Pacific countries to safeguard energy security, promote equitable growth, and advance the green transition.