AMF’s recent event brought together industry leaders to discuss the ways silicon photonics (SiPh) is maturing from a promising technology into a foundational standard for the next-generation high-speed data infrastructure. The summit’s central theme was SiPh at scale, and it provided a strategic roadmap to advance SiPh technology to meet rapidly accelerating production and unprecedented performance demands. AMF, led by a management team including CTO Patrick Lo Guo Qiang, organized the summit to showcase the collaborative ecosystem driving innovation.
AMF keynote: Scaling a $37B opportunity
The summit began with an introduction from AMF leadership and emphasized the seismic shift occurring within the market. The sheer growth of artificial intelligence (AI) workloads is driving demand for high-speed transceivers and rapidly expanding the total addressable market (TAM) for SiPh capabilities, including LiDAR and sensing. The overall optical components market is projected to reach $37B by 2029, growing at a 19% CAGR.
As a dedicated SiPh foundry, AMF is strategically positioned to capture this growth, with anticipated revenue growth of 82% from FY24 to FY25, driven by 400Gbe and 800Gbe. AMF offers core technology optimized for data centers, telecom, and sensing, with services ranging from high-volume manufacturing to multi-project wafer (MPW) shuttles—including the launch of new high-performance (HP) MPW runs, which started in September 2025.
Session 1: The AI computing tsunami and SiPh dominance
Carol Cao, senior analyst at LightCounting, delivered the opening guest presentation and described how AI computing is redefining network demands.
Unprecedented requirements. AI computing clusters require approximately 3x the quantity of optical components compared to general computing. Furthermore, AI requires dramatically higher bandwidth (>2x) and critically, low power consumption and ultralow latency.
Scale-up and scale-out. The expansion of AI clusters is astonishing, with typical large clusters jumping from 4,000 graphics processing units (GPUs) in 2022 to 100,000 in 2024. Projections for 2025 reach a colossal one million GPUs. This growth is driving new market opportunities for SiPh, especially for scale-up scenarios that require nanosecond latency.1
A technology shift to copackaged optics (CPO) and linear pluggable optics (LPO). To handle these demands, the industry is accelerating the adoption of CPO and LPO). CPO, which offers higher bandwidth density and lower latency than traditional pluggable modules, is rapidly maturing and is demonstrated by products like Broadcom’s Tomahawk 6 102.4-Tbps switch.
SiPh is key. Ms. Cao noted that SiPh is overcoming early debates about its future. Due to the rapid growth in 400G and 800G demand in 2024, the penetration rate of SiPh modules surged. SiPh offers easier capacity increases, stable performance, and a clear, reducible cost advantage. The conclusion of the market analysis is clear: SiPh will be the dominant technology for LPO and CPO.