GlobalFoundries, Marvell Expand AI Chip Capacity

According to GlobalFoundries and Marvell, the two companies have expanded their long-term partnership to increase production capacity for semiconductor technologies used in high-speed optical connectivity for artificial intelligence (AI) data centers. The announcement sent GlobalFoundries shares up more than 5% during trading, while Marvell shares gained 6.5%.

Under the multi-year agreement, GlobalFoundries will expand its silicon germanium (SiGe) process capacity at its manufacturing facility in Burlington, Vermont. SiGe technology can improve the speed and energy efficiency of semiconductor devices and is widely used in high-speed optical networking applications.

The rapid growth of generative AI models is driving data centers to handle increasingly large volumes of data between servers and computing clusters. This trend is increasing demand for high-speed optical connectivity. Compared with traditional electrical signals, optical transmission can provide higher bandwidth with lower energy consumption, making optical connectivity an increasingly important part of AI infrastructure.

Robb Johnson, Vice President of Foundry Technology at Marvell, said the expanded partnership with GlobalFoundries will help ensure access to sufficient SiGe technology and manufacturing capacity to support the significant growth the company expects in the future. The additional capacity will support Marvell's advanced optical networking products, including pluggable optical transceivers, near-package optics (NPO), and co-packaged optics (CPO).

These optical technologies can reduce transmission distances, improve data transfer efficiency, and lower power consumption. As AI computing workloads continue to scale, such capabilities are becoming increasingly important for the development of large-scale AI data centers.

The expanded partnership also highlights how growing AI infrastructure demand is extending beyond graphics processing units (GPUs) into data center networking and optical communications. As AI workloads expand, data centers require not only more powerful computing chips but also high-speed switches, optical components, and advanced semiconductor manufacturing technologies to support faster and more energy-efficient data transmission.