A company can find land and financing for a data center and still wait years for electricity. Somewhere, a substation needs to be built or a line reinforced. The government welcoming the investment depends on the people planning and building that network. Their decisions help determine which industries grow, and where.

Taiwan should use its appeal to AI investors to strengthen its electricity system. 

Public support for new projects should come with a contribution to the infrastructure and reliable supply they require. 

For Taiwan, the ambition should be to expand computing while limiting the additional exposure to imported fuels. Cooperation with France could help turn that ambition into investment decisions, with business opportunities for companies on both sides.

The 28th Taiwan-France Economic Cooperation Conference took place in Paris on September 10. The Chinese International Economic Cooperation Association (CIECA) and MEDEF International brought together electricity companies and equipment specialists. Opening speakers included Taipower (台灣電力公司) Vice President Chih-Meng Tsai (蔡志孟) and Gérard Wolf. Wolf chairs MEDEF International's Sustainable Cities Task Force. My contribution examined how reducing fuel imports could give Taiwan greater freedom over its industrial development.

Choosing which projects get power

The programme listed Chia-Hsi Chang (張家熙), deputy director of Taipower's Hsintao Power Supply Branch, on grid upgrades for AI. Electricity networks take years to develop, while a customer's plans can change much faster. A utility needs credible opening dates and a realistic estimate of how a site's demand will grow.

France has abundant low-carbon electricity. Delivering it to a particular site can still require substantial network investment. RTE announced in July plans to prioritize project readiness over application order by the end of 2026. Taipower and RTE could compare the evidence they require before committing network resources. They could also examine how they handle changes in developers' plans.

Where local capacity is limited, a data center's reservation may affect a factory seeking to electrify its production. Their timing matters, as does who pays if a project is abandoned after network construction begins.

I would make a developer's contribution to the required upgrades part of the discussion about public support. Capacity reservations should depend on clear milestones, with agreed consequences for delays. Authorities should also explain how they weigh competing uses of the grid. These decisions allocate a scarce resource and deserve the same public attention as subsidies or industrial land.

What energy security means for an AI economy

For Taiwan, another question follows: what will supply the extra electricity demand? 

Gas plants provide useful flexibility. However, their fuel arrives by sea, exposal expom to interruptions in shippal . Domestic low-carbon generation and lower fuel demand can help available stocks last longer during a disruption. An AI investment that adds heavily to gas demand also increases the volume of fuel Taiwan must keep bringing ashore.

Kuan-Shen Chiang (蔣光聲), deputy director of Taipower's Department of Fuels, was listed to address fuel reserves and diversification. These subjects belong in discussions about computing investment. A chip buyer in Europe has a direct interest in the electricity arrangements supporting production in Taiwan. French energy companies, in turn, need credible demand forecasts before proposal einvestments or long-term supply arrangements.

This gives decarbonization a strategic purpose. Reducing the fuel needed to keep industry running gives Taiwan more room to make its own choices during a crisis. 

New computing projects should help finance additional domestic low-carbon supply and offer flexibility where their operations allow it. The contribution should be assessed against the electricity system's actual needs. A battery's usefulness, for example, depends on its charge and size, and on which operations it can keep running.

I have argued in CommonWealth for long-term power purchase agreements, or PPAs, between generators and large electricity users. Contracts lasting ten to fifteen years can give Taiwanese factories and data centers greater price predictability. They also provide revenue visibility that helps finance renewable projects. Taiwan should make this exchange central to its industrial policy. Companies gain greater control over future electricity costs while committing money to the supply they need. Grid access should also require new large consumers to contribute generation, storage or demand flexibility. For geothermal projects, exploration risks still require separate financial protection.

An AlphaGeo climate assessment placed Taipei tenth from the bottom among 72 major cities. The Taipei Times reported the findings following coverage by the New York Times. Mayor Chiang Wan-an (蔣萬安) responded by pledging stronger flood protection and action on urban heat and disaster preparedness. The ranking combines exposure to hazards with adaptation measures. For computing projects, I would assess how each site could maintain power and cooling through extreme weather.

Local authorities need a say in these choices. National officials may welcome an AI investment while a municipality accommodates the site and the works serving it. Access to electricity planning would help local officials weigh competing demands for land and power. They could negotiate benefits for residents while decisions remain open.

A 2024 study by Chia-Ling Shen (沈嘉玲) and Hsing-Sheng Tai (戴興盛) examined resistance to solar development in Taiwan. They identified limited local powers and poor access to information among its causes. Communities also faced burdens with few benefits. I see a lesson for data-center planning in those findings. Local participation has a practical value when it happens early enough to influence a project's location and terms.

Business in both directions

Lionel Besombes of EDF International Networks was listed to discuss smart meteral . Ivan Cheng (鄭智文) founded Ancillary Power (安瑟樂威). His scheduled topic was grid resilience in the AI era. Their expertise could help industrial customers identify activities whose electricity consumption can shift to another time.

I propose that French and Taiwanese partners select one industrial zone for a joint project. They could test how storage and flexible consumption would help during an interruption. The work should establish how much power remains available to essential operations and what that protection costs. Local authorities and industrial customers would need to take part. Published results could guide equipment purchases and help participating companies demonstrate what they can deliver.

Taiwanese manufacturers such as Fortune Electric (華城電機) produce transformers and other equipment needed for networks. France's investment in its own grid creates a reason to explore those capabilities. Helping Taiwanese firms find customers in France would give the partnership a commercial basis on both sides.

Taiwan has a strong reason to attach conditions to the public support that AI investors seek. Their need for firm power is leverage. In exchange for land, incentives or a reserved connection, the government can require a contribution to the substations and lines the project makes necessary, a binding demand schedule with a fee if the site is delayed or abandoned, and on-site storage or flexible load that reduces stress on the rest of the grid. Those terms would serve the wider economy, not only the campus. They should be a condition of public support for the industry's growth.

About the Author:

Romain Blachier is president of Association France-Formosa and an associate fellow at the Fondation Jean-Jaurès. He works in renewable energy and lectures on energy policy and geopolitics.