Written by Scott M. Moore.
Image credit: damage of the power by H.T. Yu/ Flickr, license: CC BY-SA 2.0.
Taiwan’s dominant position in the manufacture of advanced semiconductor chips makes its energy system of global, and arguably strategic, importance. Many commentators believe that this dominant position acts as an insurance policy against conflict with mainland China, captured in the phrase “silicon shield.” Given the importance of these advanced semiconductors to the global artificial intelligence (AI) buildout, and Taiwan’s own booming data centre infrastructure, there is an equally important question of whether Taiwan’s energy system supports the notion of an “AI shield.”
Fortunately, Taiwan’s total energy generation capacity is adequate in the aggregate. Even so, Taiwan’s energy system is subject to several fragilities, including an outdated power grid, an overwhelmed grid operator, and the risk of an energy supply disruption in the event of a conflict with mainland China. Moreover, Taiwan’s energy system is increasingly politicised by concerns over energy affordability. The opposition-controlled Legislative Yuan voted in August to hold a referendum on restarting nuclear power, scheduled to be held on November 28, despite opposition from the ruling Democratic Progressive Party.
Taiwan’s energy demand has been reshaped by a combination of factors, including rapid economic growth and AI data centre demand. The Ministry of Economic Affairs revised its electricity demand forecast upward from 1.7% to 2.5% annually for the 2026-2035 period, with data centres alone expected to add up to 4 GW of power load annually. At the same time, these demand drivers have so far fallen short of placing unmanageable strain on Taiwan’s energy grid. In 2025, GDP grew by nearly 9% even as electricity consumption registered a small decrease.
The real challenge for Taiwan’s energy system is less aggregate demand than distributional shifts. AI data centre demand in particular presents a distinctive load profile that is essentially uniform around the clock and throughout the year. This presents an operational challenge, since most energy grids are built to handle variable loads that peak mostly during the day and vary across seasons. Taiwan’s grid already operates with a thin night-time power reserve, meaning that continuous load could well stretch it unsustainably. Moreover, it is not just the number of data centres that matters for the grid, but also where they are built in relation to existing grid infrastructure. Data centres can be built in a matter of months, whereas installing new transmission lines or substations can take years, making it hard for grid operators to keep up with the pace of data centre construction.
Moreover, while Taiwan’s energy system has so far met the demands associated with AI, its continued ability to do so is far from certain. The Hormuz energy shock exposed Taiwan’s continued dependence on imported liquefied natural gas, prices of which are likely to remain high for some time. The pace of renewables deployment, intended to eventually replace gas- and coal-fired power generation, meanwhile has lagged due to high costs, permitting issues, and regulatory challenges. Solar deployment fell short of a 20 GW target set by the government to be achieved by the end of 2025.
Taiwan’s energy system also requires significant upgrades to continue to meet AI-related demand. One priority is storage. Modest grid storage has been added in recent years, but meeting demand will require on-site storage for many data centres and large industrial users. Fully integrating renewables, meanwhile, will require more ambitious investments like the pumped hydropower facilities now in the planning phase. A second priority is grid resilience. Taiwan’s energy grid is island-wide and fully integrated, but generation is concentrated in the south while AI-related load exists primarily in the north, meaning that any interruption to transmission risks cutting the energy grid in half.
These grid modernisation challenges are manageable but depend on reforming a capable but fiscally precarious state utility in the form of Taipower. The firm posted a significant profit in 2025 after three consecutive years of losses, but also carries over NT$350 billion in accumulated losses. The Hormuz shock resulted in a loss of over NT$25 billion in the first half of 2026 alone. Meanwhile, energy finance has become an intensely polarised partisan issue. An NT$250 billion infusion of public funds to Taipower proposed in 2022 has so far been resisted by the opposition party.
Perhaps the biggest political issue related to Taiwan’s energy system is affordability. As in other countries, the energy-related demands of the AI buildout have focused public concern on energy prices. In response, the Lai Ching-te Administration has taken several steps intended to increase energy supply and lower prices. These include proposing the restart of Taiwan’s Maanshan nuclear power plant and a US$13 billion infusion into Taipower announced in October 2026 intended to prevent the utility from raising consumer energy prices. If these measures fail to ease voter concerns, future administrations may face growing pressure to restrict data centre development, potentially hobbling Taiwan’s AI buildout.
Ultimately, however, the validity of the AI shield concept from an energy point of view depends on the prospect of conflict with mainland China. Of course, if either the silicon or AI shield concepts have any currency, this prospect should be remote. But several scenarios short of violent conflict could severely hobble Taiwan’s energy system and AI economy. These include major cyberattacks that could damage Taiwan’s fragile power grid, or a blockade that could sharply restrict the seaborne fuel imports on which Taiwan’s electricity generation depends. Such actions could be calibrated to inflict pain on Taiwan’s economy without precipitating a collapse of its AI or other industries.
To its credit, Taiwan’s energy system has so far passed the test not only of building out the island’s AI economy, but also supplying key inputs for the global AI buildout. Yet several challenges loom, among them an outdated grid, increasingly contentious politics of energy affordability, and geopolitical tension. If these challenges can be overcome, Taiwan’s AI shield could become a viable concept. But at a minimum, Taiwan’s policymakers will need to significantly increase generation from renewable sources and lower energy prices to bring this concept to fruition. Given the strategic importance of the AI shield, whether they do so is of truly global concern.
Scott M. Moore is a Practice Professor of Political Science, Faculty Fellow of the Kleinman Center for Energy Policy, and Senior Fellow of Perry World House at the University of Pennsylvania as well as Non-Resident Senior Fellow at the Brookings Institution. His research interests include energy security and geopolitics, and he founded the Penn Global Climate Security and Geopolitics Project. Along with his colleague Thomas Shattuck, he has taught an undergraduate course on Taiwan and energy security. He previously served at the U.S. Departments of State and Defense as well as at the World Bank.
This article was published as part of the special issue on AI Shield and Energy.