ams OSRAM Unveiled Micro-VCSEL Arrays for AI Infrastructure
The company debuted an 850nm thin-film platform designed to enhance power efficiency in large-scale AI data clusters.
Updated on Sept. 27, 2026 in Semiconductors
Live Poll
Do you trust that new corporate computing innovations will lead to more reliable digital infrastructure?
ams OSRAM has unveiled a new 850nm thin-film micro-VCSEL platform engineered to improve the performance of AI interconnects. The system utilizes a parallel optical architecture to boost bandwidth while maintaining low power requirements.
Why it matters
As modern AI training and inference clusters demand massive data throughput, traditional serial channels are becoming less efficient. This wide-and-slow optical approach offers a way to reduce system complexity and energy usage in high-density computing environments.
The platform features a dense 25µm-pitch array configuration that achieved error-free 32Gb/s NRZ operation. Reliability testing confirmed the arrays remained operational without failure after 2,000 hours of stress testing.
The players
ams OSRAM
An Austria-based global leader in optical solutions that develops sensor and emitter technologies for automotive, industrial, and consumer markets.
BizLink
An interconnect solutions provider that collaborated with ams OSRAM to demonstrate a multi-core fiber connectivity solution.
The details
The technology integrates micro-VCSEL arrays onto silicon TSV substrates, combining CMOS integration with wafer-level manufacturing. By utilizing a highly parallel optical architecture, the system provides a more efficient alternative to conventional data transmission methods.
Timeline
September 21-23, 2026: ams OSRAM demonstrated the technology at ECOC 2026 in Málaga.
Roadmap
The shift toward this optical architecture reflects a broader move in the semiconductor industry to move beyond the limitations of standard serial electrical signaling. This positions ams OSRAM to capture market share as AI infrastructure providers seek to minimize energy waste in massive data centers.
For developers and hardware architects, this technology offers a path toward more efficient AI model training and data processing. Users may eventually see improvements in the performance and energy costs associated with cloud-based AI services.
The takeaway
The move toward parallel optical architectures is likely to be a defining trend for high-bandwidth AI hardware in the coming years. Developers should watch for future advancements in 3D photonics stacks as the industry seeks to optimize power consumption in data centers.
Further reading
Learn more about the latest innovations in high-performance components in the Semiconductors section.
Live Poll
Do you trust that new corporate computing innovations will lead to more reliable digital infrastructure?
