SK hynix Tech Note Explains Silicon Photonics and CPO
Understand the story
On October 29, 2026, SK hynix published a [Tech Note] article on its technical column, featuring a guest piece by Professor Kim Yong-hyun of Hanyang University ERICA that analyzes data transmission bottlenecks in AI factories. The article notes that as GPU compute speeds continue to rise, traditional copper interconnects face problems such as signal loss, increased power consumption, and heat in high-speed, long-distance transmission scenarios; by contrast, optical interconnects are increasingly advantageous thanks to WDM (wavelength division multiplexing) technology that transmits multiple wavelengths over a single fiber. The article explains the principles of silicon photonics—integrating optical waveguides, modulators, and photodetectors onto chips—to gradually build an "optical highway," addressing the problem of tens of thousands of GPUs still waiting for data during computation. At present, this is solely a first-hand technical analysis from SK hynix's memory tech note column, with no follow-up industry adoption or new developments reported.
AI overview of reporting · Updated 2 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- SK hynix Memory Technology[Tech Note] Why does AI still have to wait with tens of thousands of GPUs? How silicon photonics and CPO are building the ‘highways of light’
In SK hynix's [Tech Note] column, Professor Kim Yong-hyun of Hanyang University ERICA analyzes the data transfer bottleneck in AI factories. As GPU compute speed improves, copper interconnects face signal loss, power consumption, and heat issues at high speeds and long distances, while optical interconnects gain an edge through multi-wavelength transmission over a single fiber (WDM). The article explains how silicon photonics integrates optical waveguides, modulators, and photodetectors onto chips, gradually building 'highways of light'.
Story heat trend
Not enough continuous observations to show a trend yet.