GlobalFoundries and Marvell Technology are expanding their manufacturing agreement as the rapid buildout of artificial intelligence data centers increases demand for faster and more energy-efficient connectivity. The agreement highlights a less visible but increasingly important part of the AI infrastructure investment cycle: the semiconductor technology required to move enormous volumes of data between computing systems.

AI Data Centers Increase Demand for Optical Connectivity

The expanded multi-year agreement will increase production capacity for silicon germanium, or SiGe, technology at GlobalFoundries’ facility in Burlington, Vermont. SiGe enables semiconductor devices designed for high-speed and energy-efficient applications, making the technology increasingly relevant as data-center operators expand AI computing capacity.

The growth of AI workloads is increasing the amount of information that must move between processors, memory and other components. As data volumes rise, conventional electrical connections face greater performance and energy constraints, increasing the importance of high-speed optical networking. The GlobalFoundries-Marvell agreement therefore addresses a critical infrastructure requirement beyond the AI processors themselves.

Marvell Expands Access to Critical Manufacturing Capacity

For Marvell, the agreement provides additional manufacturing capacity to support its portfolio of advanced optical networking products. The companies said the expanded production will help Marvell meet demand for pluggable optical transceivers, near-packaged optics and co-packaged optics, technologies designed to improve how data moves through increasingly complex data-center architectures.

Marvell Vice President of Foundry Technology Robb Johnson said the expanded collaboration would help ensure access to both SiGe technology and manufacturing capacity as the company anticipates continued growth. Securing production capacity is becoming increasingly important across the semiconductor industry as AI infrastructure investment places pressure on manufacturing and component supply chains.

Investors Respond to the AI Infrastructure Signal

The announcement produced an immediate positive market response. GlobalFoundries shares rose approximately 4% in morning trading, while Marvell shares gained 6.3%. The moves suggest investors were focused on the potential implications of stronger demand for AI-related connectivity products and the companies’ positioning within the expanding data-center supply chain.

The market reaction also reflects the broader shift in how investors assess AI infrastructure. The investment cycle is no longer limited to processors and memory. Networking, optical communications, power systems, cooling and semiconductor manufacturing capacity are increasingly important components of the infrastructure required to support large-scale AI deployment.

Capacity Expansion Points to Longer-Term Infrastructure Demand

The agreement comes as semiconductor manufacturers face a global scramble to expand capacity amid strong AI-related computing demand. Recent industry developments have included efforts to diversify semiconductor production geographically and increase access to advanced manufacturing capabilities, highlighting the strategic importance of reliable supply chains.

For GlobalFoundries, the expanded Marvell relationship strengthens the role of its Vermont manufacturing operations in the AI supply chain. For Marvell, additional SiGe capacity can provide greater support for customers deploying increasingly bandwidth-intensive data-center systems. The economic significance will ultimately depend on how quickly AI infrastructure spending translates into sustained demand for optical networking equipment.

Looking ahead, investors will monitor AI data-center construction, optical networking demand, semiconductor capacity utilization and Marvell’s customer commitments for evidence of how durable the current expansion will be. The agreement also underscores a broader trend: as AI systems become larger and more interconnected, the ability to move data efficiently may become as important to infrastructure economics as the computing power generating that data.