When Nvidia CEO Jensen Huang expressed concern about Taiwan's electricity supply, President Lai Ching-te responded by inviting him to tour a power plant. That response reveals everything about the gap between politicians and business leaders. A facility visit adds not a single kilowatt to Taiwan's grid. And Lai's accompanying claim — that Taiwan's operating reserve rate stands at 20% — is, at best, a serious misstatement.
Jensen Huang Has Said It Repeatedly. The Government Keeps Not Listening.
Huang has visited Taiwan multiple times for Computex and other industry events, and on each occasion he has warned that Taiwan needs more energy to meet the demands of the AI era. When asked whether he believed the government's assurance that power supplies would remain stable through 2034, his one-word answer — "Maybe" — conveyed unmistakable skepticism.
Lai responded in a media interview by welcoming Huang to tour Taiwan's power plants on his next visit and suggesting a dinner with Taiwan Power Company (Taipower) Chairman Tseng Wen-sheng. Huang replied diplomatically that visiting a power plant alongside President Lai and Chairman Tseng would be a great honor — and then, once again, said Taiwan needs more energy.
This exchange will go nowhere useful. Even if Huang tours a Taiwanese power plant and sits down for dinner with Taipower's chairman, none of it will change Taiwan's energy policy or add a single unit of generating capacity. It is, in plain terms, a photo opportunity — one in which the world's most sought-after figure in the AI industry would be lending his star power to a political performance. Whether Huang is willing to play along remains to be seen. What is certain is that factory tours, regardless of how senior the guests or how frequently they are arranged, contribute nothing to actual power supply.
The Real Problem Is an Energy Policy Built on a Flawed Foundation
If the Lai administration genuinely wants Jensen Huang and Taiwan's business community to have confidence in the island's power supply, the answer is not study tours and ministerial dinners. It is a serious rethink of an energy policy that has been going wrong since 2016.
When the Democratic Progressive Party came to power that year, it committed to phasing out all nuclear energy and replacing it with wind and solar, while sharply increasing the share of natural gas generation and reducing coal. The policy contained a fundamental error from the outset: even if every planned target were met perfectly, thermal power would still account for 80% of Taiwan's electricity mix — meaning higher carbon emissions and worse air quality. And if renewables underperform, the thermal share climbs to 85% or beyond 90%.
That is precisely what has happened. Taiwan's geography is poorly suited to large-scale renewable deployment. The island is small and densely populated, leaving little room for utility-scale solar farms. The Taiwan Strait is largely calm in summer — the peak demand season — which undermines the reliability of offshore wind. Add permitting delays and administrative bottlenecks, and it becomes clear why a growing number of renewable energy projects have stalled or been abandoned over the past two years, with foreign developers increasingly exiting the market.
Demand Forecasts Are Too Conservative — And the Government Knows It
Meanwhile, the global AI boom and Taiwan's deepening commitment to semiconductor manufacturing have driven electricity demand sharply higher. Power consumption rose nearly 3% in 2024. It dipped 1.1% last year, largely due to reduced demand from traditional industries — and on that basis, Taipower revised its average annual demand growth forecast for the decade ahead (2025–2034) down to just 1.7%.
The Ministry of Economic Affairs' assurance that supply will remain adequate through 2034 rests entirely on this projection. It is almost certainly too optimistic. AI-driven electricity demand will not be confined to data centers. It will ripple through entire supply chains and into everyday computing, as more users engage with AI applications more frequently across more devices. That trend moves in only one direction.
Taiwan's semiconductor industry compounds the challenge. TSMC alone already accounts for more than 10% of the island's total power consumption, and that share will keep growing as process nodes advance. There is no exit from this path — the semiconductor commitment is irreversible.
The official rationale for downgrading the demand forecast to 1.7% also cites projected gains from "deep energy conservation" policies. This is not a credible basis for planning. Conservation yields diminishing returns by definition: the marginal benefit of each additional efficiency measure shrinks as the easier gains are exhausted. Decades of energy-saving campaigns have never prevented long-run growth in total consumption. Price signals, not policy slogans, drive conservation behavior — and even then, the effect has limits.
Independent analysts and academic researchers estimate that realistic annual demand growth over the next decade is closer to 2.8–3%. The official figure of 1.7% is a significant underestimate. Revising the forecast downward so sharply on the basis of a single year of negative growth — 2025 — is questionable methodology. It is no wonder Jensen Huang keeps reminding Taiwanese authorities to secure more energy. He can read a demand curve.
Lai's 20% Reserve Claim Doesn't Match Taipower's Own Data
In an effort to project confidence about Taiwan's power supply, President Lai stated that Taipower's current operating reserve rate is approximately 20% — implying that for every 100 units of electricity generated, 20 go unused. This claim does not hold up to scrutiny.
Anyone who regularly checks Taipower's real-time website, which publishes daily operating reserve figures, will rarely if ever see a number near 20%. Single-digit reserve rates are the norm. On summer peak days, the rate frequently drops below 5%. A sustained 20% operating reserve simply does not reflect the reality that Taipower's own data shows.
Lai may have conflated "operating reserve", a real-time measure of immediately available surplus capacity — with "installed reserve," a broader planning metric that includes capacity not necessarily available on demand. The two figures measure different things. Using the larger number to characterize real-time supply adequacy is misleading.
Restarting Nuclear Power Would Add 60 Billion kWh. A Plant Tour Adds Zero.
If the Lai administration genuinely wants to give Jensen Huang and the broader business community confidence in Taiwan's energy future, the path forward is not a guided tour of a generating facility. It is a substantive revision of energy policy and an acceleration of the administrative steps needed to restart nuclear power.
Were the government to commit to returning Nuclear Power Plants 1, 2, and 3 to operation within a defined timeframe — and to unblocking the long-stalled Nuclear Power Plant 4 — Taiwan could add approximately 60 billion kilowatt-hours of annual generating capacity. Northern Taiwan's chronic power shortfall would ease substantially.
That outcome is tangible, measurable, and consequential. An invitation to tour a power plant is none of those things. （Related： Unitree Robots Dance on America's Got Talent as Chinese Humanoid Maker Eyes Shanghai IPO ｜ Latest ）
