Silicon Photonics Enters Mass Production

STMicroelectronics announced it's ramping up high-volume production of its silicon photonics platform, targeting AI infrastructure—this tech uses optical, not electrical, signals to boost bandwidth and cut latency in data centers. This could enable direct, high-bandwidth connections between FPGAs, GPUs, and CPUs, reducing serialization bottlenecks. Early adopters could gain a performance edge as AI gets more embedded in trading.

STMicro's move comes as the data center optics market is projected to exceed $34 billion by 2030, growing at a 17% CAGR from $15.5 billion in 2025. Silicon photonics-based transceivers are expected to account for 76% of this market by 2030, up from 43% in 2025. This shift is driven by the increasing demands of AI, cloud computing, and 5G, which require faster, more energy-efficient data transmission. Silicon photonics uses light instead of electricity to transmit data, offering significant advantages in speed and power consumption compared to traditional copper interconnects. Optical interconnects consume 0.05 to 0.2 picojoules per bit of data transmitted, a fraction of the energy required by electrical interconnects. This technology is particularly well-suited for high-speed, short-reach data center interconnects at 400G and above. The global silicon photonics market was valued at $2.16 billion in 2024 and is projected to reach $9.65 billion by 2030, representing a CAGR of 29.5%. Other projections estimate the market could reach $10.33 billion by 2034, $13.18 billion by 2031, or even $17.8 billion by 2035. North America currently dominates the market, holding around 36.5% of the revenue share in 2024. STMicroelectronics' PIC100 platform uses 300mm manufacturing, and the company plans to more than quadruple production capacity by 2027, with further expansion in 2028. The roadmap includes PIC100 TSV (through-silicon via) technology to increase optical density and system thermal efficiency. This technology enables tighter integration of electronic and photonic components. Hyperscale cloud providers are key drivers, replacing copper links with 400 Gbps and 800 Gbps optical lanes to manage GPU cluster latency. Co-packaged optics (CPO), where photonics are integrated directly with processors and memory, are expected to contribute over $9 billion in revenue by 2030. The transition to co-packaged optics can cut switch power draw by about 30%. While Intel has been a pioneer in silicon photonics, TSMC is making strides in CPO technology and plans to mass-produce semiconductors using CPO by 2026. Nvidia is also investing heavily, committing $2 billion to expand research and manufacturing in advanced optical networking technologies. This investment will support companies like Lumentum Holdings to expand R&D and build additional U.S. manufacturing capacity. Silicon photonics faces challenges such as integrating on-chip lasers and managing yields, along with high development and production costs. However, government initiatives like the U.S. CHIPS Act and the EU's Horizon Europe program are providing funding to boost manufacturing and research. Silicon photonics offers approximately 40% lower power consumption than traditional optical modules. Ultimately, the ongoing evolution of silicon photonics promises to drive greater efficiency and performance in data centers, supporting the ever-growing demand for data in our interconnected world. The technology's ability to address the power and bandwidth limitations of traditional copper interconnects makes it a critical enabler for future AI infrastructure.

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.