GlobalFoundries and data infrastructure solutions leader Marvell Technology, Inc. have finalized an expanded multi-year manufacturing agreement. The strategic extension will significantly scale production capacity for GlobalFoundries’ proprietary silicon germanium (SiGe) process technology at its fabrication facility in Burlington, Vermont, directly addressing surging global demand for high-speed optical connectivity architectures.
Addressing AI Data Center Bottlenecks Through Advanced Optics
As modern artificial intelligence platforms evolve into massive, distributed computing networks, overall system performance increasingly hinges on the bandwidth and latency of high-speed interconnects that move data between accelerator clusters. To prevent data bottlenecks as AI models scale, hyper-scalers and chip designers are prioritizing ultra-efficient optical interconnect architectures.
Manufactured out of its Burlington, Vermont facility, GlobalFoundries’ SiGe process platform delivers high-frequency analog performance engineered specifically for next-generation optical connectivity. The technology provides native support for 200G-per-lane optical throughput, powering critical hardware applications such as pluggable optical transceivers, Near-Packaged Optics (NPO), and Co-Packaged Optics (CPO). By monolithically co-integrating silicon germanium with silicon photonics and advanced heterogeneous packaging, the platform offers the thermal efficiency, signal integrity, and high-density bandwidth required to prevent data bottlenecks in modern AI compute infrastructure.
Also Read: Vantor Partners UK Ministry of Defence to Launch UK Defence Intelligence Innovation Cluster
The expanded manufacturing agreement builds upon an established production baseline to deliver the long-term foundry capacity needed to support Marvell’s scaling product lineup. GlobalFoundries’ SiGe process portfolio delivers high-frequency signal integrity, thermal tolerance, and power efficiency required for next-generation optical networking hardware. Capable of supporting 200G-per-lane connectivity today, the SiGe technology roadmap is engineered to accommodate higher-density, next-generation bandwidth standards.
With over a decade of high-volume SiGe manufacturing, GlobalFoundries continues to refine its heterogeneous integration stack, combining SiGe components, silicon photonics platforms, and advanced packaging to enable emerging Near-Packaged Optics (NPO) and Co-packaged Optics (CPO) deployments.
“AI is fundamentally changing the way data centers are built, and connectivity is becoming increasingly critical to unlocking the performance of these systems,” said Robb Johnson, Vice President, Foundry Technology, Marvell. “As the industry moves toward higher-bandwidth optical architectures, including NPO and CPO, Marvell is investing aggressively to lead this transition. Our expanded collaboration with GF will help ensure we have the SiGe technology and manufacturing capacity to support the significant growth we see ahead.”
“SiGe is where GF’s decades of high-performance radio frequency and analog leadership translate directly into breakthrough optical performance,” said Shankaran Janardhanan, senior vice president of Photonics & RF businesses at GF. “This agreement reinforces our long-standing relationship with Marvell and reflects the critical role SiGe technology plays in enabling high-speed optical networking for AI infrastructure. We will continue investing in our technology roadmap and manufacturing capabilities both in the U.S. and around the world to support our customers’ future growth.”
Strategic Market Outlook
Extending a successful R&D and manufacturing relationship spanning more than ten years, the agreement highlights the pivotal role optical networking technologies play in enabling hyperscale AI compute environments. By securing domestic U.S. manufacturing capacity for advanced SiGe wafers, Marvell and GlobalFoundries are positioning their combined supply chain to support the next era of high-speed, energy-efficient cloud and AI infrastructure.



