Intel is increasing its silicon photonics capabilities as demand for high-bandwidth optical interconnects will increase throughout AI and high-performance computing (HPC). On the 2026 Open Computing Venture (OCP) World Summit, Intel claimed to be the one high-volume foundry able to delivering 300mm indium phosphide (InP) silicon photonics at scale to a variety of industries, in line with Wccftech. This expertise is meant to help next-generation near-packaged optics (NPO) and co-packaged optics (CPO) for AI and HPC purposes.
As highlighted by ijiwei, conventional InP substrates are typically restricted to 150mm (6-inch) wafers, with 3-inch and 4-inch codecs nonetheless extensively used, in comparison with the usual 300mm in silicon-based CMOS manufacturing. InP crystals are brittle, sluggish to develop, current defect management challenges, restrict laser chip energy per wafer, and sluggish capability enlargement. To deal with these limitations, Intel is integrating InP-based laser expertise into its 300mm silicon photonics manufacturing platform to enhance manufacturing effectivity and scalability.
NVIDIA, Broadcom and TSMC advance CPO commercialization
Intel’s push comes as CPO commercialization begins to realize momentum. In accordance with TrendForce, NVIDIA has begun transport its Spectrum-X CPO switches to pick out companions, whereas Broadcom continues to ship restricted shipments of its 51.2 Tb/s Bailly CPO switches. NVIDIA’s Spectrum-X CPO swap, co-developed with TSMC, makes use of the foundry’s Compact Common Photonic Engine (COUPE) expertise.
Broadcom can also be increasing its CPO portfolio. At OCP 2026, the corporate showcased its third era Tomahawk 6-Davisson CPO swap, which provides 102.4 Tb/s of switching capability and 200 Gbps per channel. In the meantime, TSMC plans to start manufacturing of its COUPE-on-substrate answer in 2026, with the objective of doubling energy effectivity and lowering latency by 10x in comparison with board-level pluggable optics.
Chinese language suppliers and cloud suppliers are additionally creating NPOs and CPOs. In accordance with ijiwei, Zhongji Innolight has established a devoted CPO manufacturing line and began producing small portions of 800G CPO optical modules. Some prospects have positioned orders, and shipments may start in 2027, with broader adoption anticipated in 2028. Chinese language cloud suppliers are additionally making progress. In September 2026, Huawei launched NPO-based Ascend 960 supernodes for large-scale AI coaching and inference.
CPO testing emerges as a serious bottleneck in mass manufacturing
Regardless of progress towards commercialization, testing stays a serious impediment to mass manufacturing of CPO. In accordance with TrendForce, optical alignment instances have been decreased to seconds or much less, however an entire check of a single photonic built-in circuit (PIC) can nonetheless take greater than 100 seconds, and the bottleneck has shifted to check pace and value. TrendForce expects 2027 to be a pivotal 12 months for CPO check system integration, and 2028 to be a important time to evaluate readiness for large-scale manufacturing.
(Photograph courtesy of Intel)
TrendForce 2027 Infrared Sensing and Optical Communication Market Developments
1. Annual report
Launch: January 1, 2027
Format: PDF/EXCEL
Language: Conventional Chinese language / English
2. Optical Communications – Quarterly Worth Report for InP Laser Diodes and InGaAs Photodiodes
a. InP board – 3 inches
b. EML-100/200G
c. CW-DFB LD-50/70/100mW
d. InGaAs Photodiode – 100/200G
e. DSP-800G/1.6T
Launch date: twentieth of February, Might, August, November 2027
Format: EXCEL
Language: Conventional Chinese language / English
|
If you need extra data, please contact us under. |
