Taiwan's AI Data Center Capacity Lags Behind China Due to Power Constraints, Semiconductor Executive Warns

Taiwan's inadequate electricity infrastructure poses a significant barrier to the island's artificial intelligence ambitions, according to a semiconductor industry executive who rated the country's power capabilities far behind those of China and the United States.

Yeh Kuo-kuang (葉國光), general manager of semiconductor firm Chinghui,outlined three critical factors for optimal AI data center operations during a television program: chip computing power, cooling capacity, and electricity supply. He warned that insufficient cooling leads to temperature increases that reduce computing performance, while reliable power supply forms the foundation for AI system operations.

In Yeh's comparative assessment of AI data center capabilities across the three regions, China emerged as the overall leader with 26 points, narrowly ahead of the United States at 25 points, while Taiwan lagged significantly at 15 points. （Related： Nvidia CEO Jensen Huang Signals Bold Expansion in Taiwan: "The More Headquarters, the Better" ｜ Latest ）

The United States scored highest in chip computing power (10 out of 10), with Taiwan following at 8 points and China at 6 points. However, China dominated in cooling capacity and electricity supply, scoring maximum points (10) in both categories. The U.S. achieved 7.5 points in each of these areas, while Taiwan managed only 5 points for cooling and a mere 2 points for electricity.

"If you want to develop AI, first you need sufficient electricity, second you need cheap electricity. Otherwise, what's the point of talking about AI every day?" Yeh said, according to the program. "I gave Taiwan's electricity a score of 2 points, and I was being generous - I originally wanted to give it 0 points."

Electricity costs vary dramatically across the three markets, according to Yeh's data. China's Xinjiang region offers the lowest rates at approximately $0.036 per kilowatt-hour during peak times and $0.033 during off-peak periods. Inner Mongolia charges roughly $0.043 per kilowatt-hour at peak and $0.035 during off-peak times.

By contrast, the United States charges between $0.14 and $0.18 per kilowatt-hour during peak periods, dropping to around $0.10 during off-peak times. Taiwan's rates range from $0.15 to $0.22 per kilowatt-hour at peak, though they fall to approximately $0.075 during off-peak periods.

Yeh emphasized nuclear power's cost advantages and reliability, pointing to the ongoing Russia-Ukraine conflict as evidence. "Nuclear power stations have never been damaged and continue generating electricity, while all other power sources have been destroyed," he said, calling electricity a crucial factor for AI's future development.

Looking ahead, Yeh noted China's commitment to achieving carbon neutrality by 2050 through wind, solar, and nuclear power sources. He predicted that China would become a clean energy powerhouse capable of driving its AI industry with abundant, exportable electricity supplies.