GlobalFoundries Signals Expanded GaN Strategy for AI Data-Center Power Demand

GlobalFoundries has reaffirmed its global manufacturing strategy and identified Taiwan as a key ecosystem for Data Center AI and Physical AI, while linking its GaN roadmap to rising power-component demand from AI data centers.

According to the company, demand from US hyperscalers continues to support local manufacturing in the United States. GF also highlighted its distributed manufacturing footprint, citing the United States and Singapore, as it pursues a broader global capacity expansion strategy.

The company said it has licensed high-voltage GaN technology following a partner’s exit from the GaN market, with the stated purpose of transitioning existing customers. GF described GaN as applicable across a voltage range from high-voltage applications of up to 800 V to low-voltage applications in the single-digit-volt range.

For data-center infrastructure, GF’s emphasis is on low-voltage GaN power devices. The company says demand in this segment is growing rapidly and that it intends to improve device reliability while expanding global manufacturing capacity. This focus is relevant to AI data centers, where rising compute loads increase the importance of efficient, reliable power delivery throughout the electrical chain.

The announcement also points to a potential shift in the power-semiconductor supply chain for AI infrastructure: demand may extend beyond high-voltage devices to low-voltage components, where on-resistance, die-area scaling and reliability are critical. GF’s geographically distributed manufacturing could be relevant for customers seeking to reduce supply-concentration risk, although this remains an inference rather than a disclosed customer strategy.

GF did not provide production-capacity figures, CAPEX, named fabs, expansion locations, implementation dates or project status for the planned global expansion.