As artificial intelligence clusters grow larger, the ability to move data efficiently between processors has become as critical as raw computing power. Current optical interconnects struggle to keep pace with the bandwidth, latency, and power demands of AI training and inference at scale. ams OSRAM's latest announcement addresses this bottleneck head-on, unveiling a breakthrough Thin-Film VCSEL platform and a connectorized multi-core fiber solution developed with ecosystem partner BizLink. The move signals a shift toward highly parallel, wide-and-slow optical architectures that could fundamentally alter how data centers are built.
The core of the announcement is a next-generation 850 nm Thin-Film microVCSEL array platform designed specifically for AI scale-up interconnects. Unlike conventional VCSELs, these devices use top-emission Thin-Film structures with dense 25 µm pitch arrays that are individually addressable. Combined with silicon through-silicon via (TSV) integration, the platform enables exceptionally high packaging density and compatibility with standard multimode fiber. Critically, ams OSRAM has demonstrated error-free 32 Gb/s NRZ operation with an energy efficiency of approximately 0.25 pJ/bit. That figure is significant because power consumption is a primary constraint in AI data centers; reducing the energy per bit directly lowers operating costs and thermal loads. Reliability testing also showed no failures after more than 2,000 hours under elevated junction temperatures and current overstress, suggesting the technology is robust enough for deployment.
The strategic importance extends beyond the component itself. ams OSRAM is not simply selling a faster VCSEL; it is building an ecosystem around a wide-and-slow architecture. In this approach, data is distributed across many lower-speed optical lanes rather than concentrated in a few high-speed channels. This reduces system complexity, lowers latency, and allows bandwidth to scale more gracefully. For AI scale-up networks, which connect accelerators within a cluster, such an architecture could deliver the low latency and high density required while keeping power in check. The partnership with BizLink, a specialist in fiber-optic connectivity and optical integration, is essential for realizing this vision. Their joint demonstrator at ECOC 2026 combines ams OSRAM's 2D addressable microVCSEL array with BizLink's fiber-array integration to create a connectorized multi-core fiber solution. This addresses a often-overlooked challenge: even the best emitters need practical, manufacturable fiber connections to be useful.
The technology builds on ams OSRAM's established Digital Photonics platform, which was previously commercialized in EVIYOS, an advanced digital lighting solution for adaptive headlamps. That heritage matters because it proves the underlying micro-emitter array, CMOS integration, and wafer-level manufacturing technologies can be produced in high volumes with consistent quality. Extending the same principles to AI infrastructure lowers execution risk and accelerates time to market. The company has also announced a confidential customer cooperation earlier this year and is collaborating with multiple partners across testing, fiber infrastructure, packaging, and system integration. This collaborative approach is designed to drive industry-wide adoption of next-generation optical interconnects.
For the AI industry, the implications are substantial. As clusters scale, the interconnect fabric increasingly determines overall system performance and efficiency. A wide-and-slow optical architecture based on addressable microVCSEL arrays could enable more sustainable and cost-effective data centers, where bandwidth grows without proportional increases in power. The connectorized multi-core fiber demonstrator, to be shown privately to selected customers at ECOC 2026, validates that this architecture is not just theoretical. Additional details on ams OSRAM's optical interconnect developments are available on the company's dedicated technology page, and the original release can be viewed on www.newmediawire.com. The company also maintains a presence on social channels such as LinkedIn and YouTube. As AI infrastructure continues its rapid expansion, ams OSRAM's wide-and-slow photonics platform offers a compelling path toward lower power, lower latency, and scalable bandwidth.
