The growing silicon photonics space just got that little bit more mature with the release of a standard aimed at ensuring consistent reliability among devices.

The Joint Electron Device Engineering Council (JEDEC), the association behind some of the most integral memory frameworks, unveiled JESD264, providing a common baseline for testing and manufacturing controls for silicon photonics.

An apparent industry first, the framework covers reliability protocols for silicon photonic dies and integrated chipsets, as well as guidelines for the component level across everything from modulators to photodiodes.

Manufacturing process controls and laser integration requirements are also covered, with the association stating its standard should help reduce deployment risk while encouraging broader adoption in data centers, telecom, and “AI networks.”

Traditional complementary metal-oxide-semiconductor (CMOS) technologies encompass a plethora of established standards given their broad adoption, like in radio frequency systems with Wi-Fi and Bluetooth transceivers. JEDEC contends, however, that the same can’t be said for silicon photonics, resulting in “uncertainty across the supply chain.”

JESD264 looks to fill the gap for manufacturers, integrators, and end users.

“As AI, cloud computing, and high-speed communications continue to drive demand for greater bandwidth and performance, industry standards become increasingly important,” JEDEC Board of Directors Chair Mian Quddus noted. “JESD264 helps establish the reliability foundation needed for broader silicon photonics adoption, accelerating innovation and strengthening alignment across the industry.”

Silicon photonics is quickly gaining intense traction from folks in the AI infrastructure space owing to its ability to usurp slow, heat-generating copper with light-based links capable of processing data at higher speeds using less power. Hype for the emerging technology has resulted in projections of a 23-fold surge in silicon photonic-specific revenue between 2024 and 2030, according to a recent China Insights Consultancy (CIC) report.

The likes of Coherent, Marvell, and Credo are all making strides to mature the technology for use in AI networks. However, rival segments like near-packaged optics (NPO) and co-packaged optics (CPO) are drawing attention owing to the fact that they are far less expensive to produce than silicon photonics.