AI summary
Summary of article:
Intel revealed its silicon photonics blueprint at OCP 2026, emphasizing that it can break the transmission limitation of copper wires and use 300mm indium phosphide and packaging integration solutions to accelerate the implementation of data center optical communications.
intel He believes that silicon photonics will be the core architecture of the next generation of high-performance computing and AI data centers, and describes silicon photonics as a “golden hammer”. He explains that the company is the only company in the industry capable of mass producing 300 mm indium phosphide (InP).
Intel unveiled its silicon photonics technology blueprint at the 2026 Open Computing Projects Summit (OCP 2026). Nick Dube, vice president of data center systems and solutions, said silicon photonics is like a golden gavel. Silicon photonics as an industry standard has been talked about for several years, and the Intel team is committed to enabling this technology to be deployed in the real world within the next two years.
He said traditional copper wire transmission is approaching physical limits in three areas, including transmission distance, bandwidth density at the edge of the package, and energy per bit. Silicon photons can solve these three major problems simultaneously.
InP is a III-V semiconductor that can generate laser light sources and can be integrated into silicon photonics platforms to provide the light source needed for light transmission. Compared to designs relying solely on external light sources, integrating laser light sources helps reduce packaging and light pass complexity. Intel said the company’s platform uses highly stable light sources and supports industry-standard external light sources, providing multi-wavelength light sources supplied by US wafer fabs while maintaining the design flexibility of different architectures.
In terms of deployment, Intel uses co-packaged optics (CPO) and near-packaged optics (NPO) to place optical components as close as possible to computing chips and network switching chips to reduce the high-speed transmission distance of electronic signals. Intel aims to reduce the per-unit transmission energy efficiency of optical communications to less than 5pJ. The semiconductor industry has always recognized that this is a significant limitation to the commercialization of optical communications. However, the above are Intel’s technical goals, and the company has not announced an actual mass production program.
Summary FAQ
-
Because Intel believes that silicon photonics can simultaneously solve the three major constraints of copper wires in transmission distance, packaging edge bandwidth density and energy per bit, and will play a fundamental role in future high-performance computing and AI data centers.
-
Intel said it is the only company in the industry that has the ability to mass produce 300 mm indium phosphide InPs, and its platform can provide a stable light source and support external light sources and multi-wavelength light sources supplied by US wafer fabs, which maintains architectural flexibility.
-
Intel uses co-packaged optical CPOs and near-packaged optical NPOs to place optical components as close as possible to the computing and switching chips, and aims to reduce the per-bit transmission energy efficiency to less than 5pJ; However, the company has not yet announced an actual mass production schedule.
