Over the past half-year, multinational companies and semiconductor suppliers have pressed the government to reduce surplus power conditions. To increase the flexibility of green power transfer and matching, Taiwan Power (Taipower) has planned to launch Sandbox 3.0 in October 2026, which market watchers expect will help address TSMC's surplus electricity issue and add flexibility to Taiwan's renewable energy market.
Unlike RE100, which calculates green power procurement on an annual basis, Google's 24/7 Carbon-Free Energy strategy aims to ensure that every hour of electricity used by data centers comes from carbon-free energy through precise matching of time and location.
Because renewable energy is intermittent, RE100 can still leave certain time periods dependent on non-green electricity, while 24/7 adds a time-based requirement that makes green power use stricter.
To cover around-the-clock power needs with green electricity, Google has adopted a diversified procurement strategy in Taiwan that includes solar, offshore wind and geothermal power. Because these three sources have different generation curves, they can complement one another at different times.
Google is reportedly not participating in Taipower's sandbox program, instead trying to avoid surplus electricity by purchasing renewables such as wind, solar and geothermal power.
As large-scale offshore wind projects come online, Google has already announced a partnership with Copenhagen Infrastructure Partners (CIP), signing the first offshore wind power purchase agreement in Taiwan and the Asia-Pacific region. The Fengmiao 1 project is scheduled to begin operations in 2027 and will provide stable power for Google's data centers and offices in Taiwan.
People familiar with the matter said Google does not currently face a surplus electricity problem and is pushing hard to raise its RE score, but it could join Taiwan Power's sandbox program if such an issue emerges in the future.
They added that data centers used only for storage have relatively stable power demand, but if they are upgraded into AI data centers with heavy computing requirements, the surplus electricity problem would become more severe.
Energy supply chain players estimate that one AI data center consumes 30-120MW of electricity, equivalent to 1.05 billion kWh a year. As computing demand rises, that figure will reach 1GW, with annual power consumption of about 8.7 billion kWh.
Article translated by Charlene Chen and edited by Jack Wu
