July 25, 2025 /SemiMedia/ — AMD CEO Lisa Su said that chips manufactured by Taiwan Semiconductor Manufacturing Co. at its Arizona plant could cost up to 20% more than those made in Taiwan, raising concerns about the financial impact of shifting chip production to the U.S.

Speaking at an AI industry event in Washington, Su stated that the cost difference for U.S.-made chips would likely range between 5% and 20%, depending on the product. Despite the added expense, she said the trade-off is necessary to strengthen the company’s chip supply chain. “We have to think about supply chain resiliency,” she said. “We learned a lot during the pandemic.”

AMD expects to receive its first batch of chips from the Arizona site by the end of this year. Su added that the plant’s chip yields are already on par with TSMC’s Taiwan fabs.

The remarks come as the U.S. pushes for more domestic chip manufacturing to reduce dependence on Asia. However, the higher costs have raised questions within the industry about long-term competitiveness.

Su also emphasized the continued demand for AI accelerators, citing major investments by OpenAI’s Sam Altman and xAI’s Elon Musk. AMD is one of the few viable challengers to Nvidia in the AI accelerator space, crucial for training and running large-scale AI models.

The event also marked the launch of the U.S. government’s “AI Action Plan.” Su praised the initiative, calling it “very actionable,” and welcomed collaboration between industry and government.

On export controls, Su noted that AMD and rivals are still navigating temporary restrictions on shipping certain AI chips to China. The duration and scope of these restrictions remain unclear. She urged U.S. policymakers to allow exports to allies in order to maintain global leadership in AI infrastructure.