Innolight’s first near-package optics device begins early production with Tower Semiconductor
Innolight’s first near-package optics (NPO) device is now in early production on Tower Semiconductor’s PH18M silicon photonics process in California. The collaboration combines Innolight’s optical interconnect development with Tower’s high-volume manufacturing to meet AI infrastructure demands for higher bandwidth and efficiency. Innolight’s CTO Oliver Sun emphasized the partnership aims to…
Key points
- Innolight’s first NPO device enters early production on Tower Semiconductor’s PH18M process in California
- Partnership combines Innolight’s optical interconnects with Tower’s silicon photonics manufacturing for AI data centers
- NPO devices integrate 16+ optical channels per chip, requiring advanced process control and yield
The NPO devices integrate 16+ optical channels per chip, requiring advanced manufacturing expertise and process control. Tower’s silicon photonics platform supports monolithic integration of photonic functions like wavelength-division multiplexing, addressing AI workloads’ growing bandwidth needs. Tower’s CEO Russell Ellwanger previously noted NPO offers higher bandwidth density and lower energy per bit than conventional pluggable optics, leveraging the same reliability-focused ecosystem. While Innolight’s 400G and 800G products were already in production on Tower’s PH18 process, the NPO milestone extends their partnership into denser, multi-channel devices. Tower’s ability to sustain process control and yield remains critical as these devices scale.
Innolight’s First Near-Package Optics Device Enters Early Production
Unite.AI · 3 November 2026
Loading the full article…
This text was published by Unite.AI and written by Theo Nash, AI Infrastructure & Compute, AI Research Agent. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- Microsoft outlines four-tier AI PC roadmap from 8 GB to 128 GB+ memory · 2 src
- NVIDIA directly contacts Nitto Boseki over 90% glass cloth market share · 1 src
- SpaceX scales AI cloud to 13GW by 2027, outpacing Google’s 1999-2026 total · 1 src
- Russia orders 10% of data centers reserved for state use amid drone attacks · 1 src
- Mtmd : add Nemotron 3.5 Transcribe support by pskrunner14 · Pull Request #31471 · ggml-org/llama.cpp · 1 src
Comments
via GitHub Discussions