From electrons to photons: Scaling optical connectivity with silicon photonics and SiGe
AI infrastructure is scaling simultaneously within the rack, across the data center and between data center campuses. That multidimensional growth is redefining the industry’s core challenge. The constraint is no longer compute alone, but the speed, efficiency and reliability of data movement at scale. As AI workloads grow more distributed and bandwidth-intensive, the industry is exceeding the practical limits of copper—accelerating the shift to optical interconnects that can deliver the performance, energy efficiency and resilience required across the architecture.
At the center of this transition are two tightly coupled technologies: silicon photonics (SiPh) and silicon germanium BiCMOS (SiGe). Together, they deliver the bandwidth density, energy efficiency and signal integrity required as optical interconnects scale from 800G to 1.6T and onward to 3.2T per optical transceiver—and toward the multi‑terabit systems now coming into view.
Silicon photonics brings optics onto a scalable silicon manufacturing foundation, while SiGe delivers the high‑speed analog and mixed‑signal electronics that make those optical systems practical and efficient. But performance at scale depends on more than the optics or electronics alone. It’s their co‑optimization across devices, electronics and packaging that turns optical innovation into deployable AI infrastructure.
GF is ramping these critical technologies today, drawing on more than a decade of sustained SiPh investment, over four decades of SiGe innovation and deep manufacturing expertise across a growing ecosystem. And further still, by combining these technologies with advanced packaging, we’re enabling a new class of optical platforms designed to meet the power, performance and reliability demands of large‑scale AI data centers.
In this issue, we explore SCALE, GF’s co-packaged optics (CPO) solution for modern AI scale‑up compute architectures, share insights from our first Optica Fellow and examine how SiPh and SiGe are being manufactured, integrated and deployed in production today.
Introducing SCALE solution: The OCI MSA-capable co-packaged optics platform, ready for AI scale-up
Modern AI data center infrastructure is now defined by scale. Rapid growth in model complexity, training data and east‑west traffic has exposed hard limits in today’s copper‑reliant frameworks, forcing the industry to rethink system efficiency at every level. Simply increasing speed is no longer enough unless bandwidth density and energy efficiency advance in parallel.
Co‑packaged optics (CPO) represent a critical breakthrough for enabling these expansive AI workloads, integrating photonic ICs (PICs) with electronic IC (EICs) into a single functional module with the power, performance and compact sizing required for next‑generation scale‑up data center architectures. At GF, our leadership in silicon photonics technologies has already demonstrated success in addressing the bandwidth, reach, density and efficiency requirements set by major industry leaders. That momentum is reflected in initiatives such as the OCI MSA, which aims to standardize CPO and accelerate the transition to optical connectivity at the CPU, GPU or AI Accelerator and Network Switch.
Our recently announced SCALE™ CPO solution brings together GF’s proven technology and manufacturing expertise to deliver a platform that is production‑ready for widespread deployment. SCALE—our Silicon Photonics Co‑packaged Advanced Light Engine—leverages both dense wavelength‑division multiplexing to significantly reduce fiber counts and increase overall bandwidth density, supporting 8λ and 16λ per fiber today. Combined with a portfolio of fully qualified photonic devices, detachable fiber‑attach solutions and integration of EICs as per our customer requirements, SCALE provides a comprehensive, deployable path to accelerating optical scale‑up interconnect adoption and keeping pace with AI‑driven demand.
Speaking at GF’s Investor Day 2026, Chief Technology Officer Gregg Bartlett made it clear that the shift from copper to photonics is no longer theoretical. The focus now is on scaling photonics fast enough to meet the demands of next‑generation infrastructure:
Advancing high-performance SiPh and SiGe for the next era of optical connectivity
Optical interconnects have become a defining constraint on system performance, and solving it requires more than photonics alone. SiPh enables data movement, while SiGe delivers the ultra‑high‑speed analog and mixed-signal electronics that drive, receive and condition those signals with the precision and efficiency modern systems demand.
The industry recognizes SiGe as a proven, high‑performance platform for addressing the RF, timing and mixed‑signal challenges emerging alongside next‑generation optical and AI infrastructure.
In this blog, GF experts outline how the convergence of SiPh, SiGe and advanced integration is shaping the future of optical modules—from pluggables to co‑packaged optics—and moving these architectures into production.
A conversation with GF’s pioneer silicon photonics leader and Optica Fellow Dr. Yusheng Bian
At 1310 nanometers—roughly the width of a single bacterium—the light pushing hundreds of gigabits per second through data server exists beyond the edge of human vision. Hitting the retina and disappearing, it carries the silent weight of our digital world across switches, storage and AI compute clusters.
Recommended by LinkedIn
That sense of imperceptibility does not apply to GF Master Inventor Dr. Yusheng Bian. The energy this Optical Engineering Ph.D. holder emits when talking about silicon photonics is unmistakable. He answers questions with his hands as much as his words and makes a habit of apologizing for getting lost in the nuances of his life’s work.
“The moment I realized the same silicon platform we used for electronics could also guide light, it completely changed how I thought about what was possible,” Bian said.
When Bian joined GF in 2017 as a silicon photonics researcher, silicon was still primarily a medium for moving electrons. While silicon photonics had been demonstrated, its nascent ability to use light—rather than electrons, at scale and with far greater efficiency—was still emerging.
Light that small requires faith. In Bian’s case, that faith has since been affirmed by the field itself, most recently with his elevation to Optica Fellow, one of the highest recognitions in photonics and optical engineering.
We sat down with Bian to discuss the evolution of silicon photonics from research to reality, why its role in AI infrastructure is only just beginning and how a different expression of waveform and frequency inspires his work. Here’s a preview of the conversation:
What are some things that you've really been focusing on in the silicon photonics space?
Yusheng: I think over the past few years, both GlobalFoundries as a team and I personally have made tremendous progress. We are now seeing AI data centers evolve to a point where traditional copper‑based electronic solutions can no longer keep pace with increasing data demands—especially for data transfer between chips and across data centers.
Because of that, silicon photonics is now widely recognized as a must‑have solution for next‑generation data transfer, replacing copper where electronic interconnects can no longer scale. It enables lower power consumption, longer transmission distances and significantly better signal integrity.
This represents a real paradigm shift—from traditional copper wiring, which everyone is familiar with, to optical fiber, and now toward integrating optical functionality directly into silicon photonics, or photonic engines, that can be placed very close to electronic systems.
Was this explosion in demand for silicon photonics driven by AI something you anticipated?
Yusheng: Even a decade ago when I joined GlobalFoundries as one of the earliest R&D engineers purely focused on silicon photonics and we launched our silicon photonics technology this was not obvious. Back then, we knew silicon photonics was coming, but across the industry there was no clear consensus that it would definitively replace copper or traditional electronic solutions.
Over the last few years, however, with the AI boom, rapidly growing data center demand and strong momentum from key ecosystem players, we’ve gained much greater confidence. Photonics is now clearly one of the essential solutions needed in data centers to address current bandwidth, power and signal integrity bottlenecks.
This wasn’t a big surprise or completely unexpected, but we are seeing strong confidence and real momentum toward photonic solutions across the board. And this extends beyond data centers. We’re also seeing adoption in areas such as photonic computing, quantum computing, sensing and other emerging applications.
Photonics may not be a ubiquitous solution yet, but it is becoming very widely adopted as a replacement—or complement—to state‑of‑the‑art electronic technologies. Over the next decade, we expect to see major advancements and extensive deployment across these fields.
Find the full conversation with Bian:
Did you know?
GF holds approximately 3,000 combined SiPh and SiGe patents
This number of combined patents is one of the largest portfolios among leading semiconductor manufacturers.
Subscribe to be notified when the next issue of GF’s Where the future takes shape is published.
Interested Topics Newsletter : I invite you to visit my new Weekly Newsletter Bulletin ✍️ https://www.linkedin.com/newsletters/the-evolving-market-news-7349475298716839936
Interested Topics Newsletter : I invite you to visit my new Weekly Newsletter Bulletin ✍️ https://www.linkedin.com/newsletters/the-evolving-market-news-7349475298716839936