Intel Scales Indium Phosphide for Silicon Photonics
The company reveals high-volume production capability for 300mm wafers to meet growing AI bandwidth demands.
Updated on Oct. 19, 2026 in Semiconductors
Live Poll
Do you believe rapid advancements in AI hardware will lead to lower technology costs for users?
Intel has announced it can now supply 300mm indium phosphide silicon photonics at scale, addressing a critical bottleneck in manufacturing. This development marks a significant shift from conventional 150mm substrate limits as the industry races to improve optical interconnects.
Why it matters
Rising demand for high-bandwidth optical interconnects in AI and high-performance computing requires larger, more efficient wafers. Because indium phosphide is brittle and prone to defects, shifting to larger substrates is essential for increasing output.
Intel utilizes 300mm wafers to house its photonics technology, effectively doubling the diameter compared to the standard 150mm limits. Testing a single photonic integrated circuit remains time-intensive, requiring over 100 seconds per unit.
The players
Intel
This multinational corporation is a leading designer and manufacturer of semiconductor chips and computer components.
NVIDIA
This technology company specializes in designing graphics processing units and hardware for artificial intelligence applications.
Broadcom
This firm develops and supplies a wide range of semiconductor and infrastructure software solutions for data centers.
Zhongji Innolight
This company focuses on the research and production of high-speed optical transceivers and interconnect modules.
Huawei
This multinational technology corporation provides telecommunications equipment and consumer electronics.
The details
Intel integrates indium phosphide-based laser technology directly onto a 300mm silicon photonics platform to overcome historical defect-control challenges. Competitors like NVIDIA and Broadcom are concurrently scaling co-packaged optics, with Broadcom showcasing switches like the 51.2 Tb/s Bailly and the 102.4 Tb/s Tomahawk 6-Davisson.
Timeline
September 2026: Huawei introduced the NPO-based Ascend 960 supernode.
October 2026: Intel announced its 300mm indium phosphide capability at OCP.
2026: TSMC plans to begin COUPE-on-substrate production.
2027: Potential shipments for Zhongji Innolight and CPO system integration.
2028: Broader adoption of CPO and mass production assessment.
The Tech Race
The transition to 300mm indium phosphide wafers represents a pivotal shift from legacy 150mm manufacturing to support the requirements of co-packaged optics technology. This capability positions Intel to compete directly with rivals by addressing the physical constraints of brittle laser materials.
The ability to produce larger, high-performance optical chips at scale will lower costs and improve bandwidth for data-intensive AI and computing services. Users can expect more efficient cloud infrastructure and faster data processing as these advanced chips reach broader commercial adoption.
The takeaway
Scaling indium phosphide wafer sizes is a technical necessity to keep pace with the massive data demands of modern artificial intelligence. As production moves toward 300mm platforms, the industry expects a transition to more reliable and higher-capacity optical connectivity by 2028.
What happens next
TSMC is scheduled to initiate production of its COUPE-on-substrate solutions in 2026, while Zhongji Innolight expects potential shipments to begin in 2027.
Further reading
Explore more developments in Semiconductors to understand the evolving landscape of high-performance chip production.
Live Poll
Do you believe rapid advancements in AI hardware will lead to lower technology costs for users?
