News
GlobalFoundries and Samsung present advanced silicon photonics research at OFC 2026
Sunday, March 15, 2026 at 08:23 PM
Technical papers presented at OFC 2026 by GlobalFoundries and Samsung suggest significant advancements in silicon photonics and optical interconnect technologies, critical for next-generation AI data center infrastructure.
Context
At the OFC 2026 conference in Los Angeles, GlobalFoundries and Samsung Electronics showcased critical advancements in silicon photonics aimed at overcoming the "bandwidth wall" in AI data centers. GlobalFoundries presented three technical papers, including collaborations with NLM Photonics on 1.6T and 3.2T Photonic Integrated Circuits (PICs) utilizing their advanced AMF process. This development is significant as silicon photonics is projected to reach a $32 billion market by 2031, with GlobalFoundries currently positioned as a top-tier foundry for these optical interconnects.
Samsung Electronics used the event to signal its aggressive roadmap to narrow the gap with TSMC, reportedly aiming for mass production of silicon photonics by 2028. This strategy is being paired with a pivot toward glass substrates, which offer better thermal alignment with silicon and enable massive "super-packages" for next-generation AI. Regarding the specific tweet attributed to @PhotonCap noting they had "finished reviewing all three $GFS papers as well as two Samsung papers," this specific statement could not be independently verified as a formal quote from a company executive or official press release, but it aligns with the heavy research presence both firms maintained at the March 2026 technical sessions.
Sources (6)
University Publications | GlobalFoundriesIntel and Samsung Pivot to Glass Substrates for the Next Era of AI ...Samsung reportedly aims to begin silicon photonics mass production in 2028Silicon Photonics In The Data Center: What A CMOS Exec Needs To KnowNLM Photonics Samples Silicon Organic Hybrid PICs Manufactured at GlobalFoundries | OFCPerformance, efficiency, and cost analysis of wafer-scale AI accelerators vs. single-chip GPUs - ScienceDirect
Related Companies
Samsung Electronics
005930