I published an op-ed in the Taipei Times, in the Editorial & Opinion section, titled “Realizing Taiwan’s AI ambitions.”

The starting point: the France-Taiwan conference of 11 September in Paris

The piece follows the Taiwan-France Economic Cooperation Conference, co-chaired in Paris by Taipower vice president Tsai Chih-meng (蔡志孟) and Gerard Wolf, chairman of MEDEF International’s Sustainable Cities Task Force. The meeting, organised by the Chinese International Economic Cooperation Association and MEDEF International with the involvement of the Taipei Representative Office in France, also heard Representative Hao Pei-chih (郝培芝). I took part in the event: the article quotes me arguing that Taiwan should look at energy security in terms of different kinds of dependence.

The argument

• AI is becoming electricity policy. The conference’s three sessions covered grid modernisation, infrastructure resilience in the AI era, and energy security and geopolitics.
• Connection requests running ahead of reality. Taipower has received 75 transmission-level connection requests from data centres, representing 4,523 MW of requested capacity by the end of July. Of the 47 requests that received Taipower’s agreement, about 40% expired before a formal application was submitted — an issue France also faces, with RTE now prioritising project maturity over the order applications entered the queue.
• Infrastructure costs that don’t disappear. A new data centre or semiconductor plant can require major upgrades to substations and transmission lines; a five-year delay does not erase those costs, especially if several very large consumers arrive in the same area.
• Two-way industrial cooperation. Fortune Electric, whose representatives I met during the conference, manufactures transformers: Taiwan is not simply a market for European energy technology, its companies can also contribute to the huge investment now required in European grids. The wider Taiwanese mission in France also met RTE, TotalEnergies and Hitachi Energy.
• Different kinds of dependence. Gas provides flexibility but requires frequent deliveries by ship; nuclear power also depends on imported fuel and services, but can be planned over much longer periods; offshore wind produces electricity locally but needs a stronger grid; geothermal can provide firm domestic generation, but first requires proving the resource — an exploration risk France has experience sharing through dedicated mechanisms that could be discussed with Taiwanese partners.

What I propose

Not to bet on a single technology to solve every problem: gas, nuclear, offshore wind, solar, storage and geothermal do not create the same vulnerabilities. Taiwan and France should compare electricity demand scenarios linked to AI and semiconductor investment, share experience on grid planning and large consumers, work on storage and flexibility, and select a small number of concrete projects where the results of cooperation could actually be measured.

For an economy betting heavily on semiconductors and AI, whether the entire electricity system — generation, fuel, grids, storage and equipment — can support that industrial strategy might become one of the most important economic questions of the decade.

→ Read the op-ed in the Taipei Times