Avicena Advances microLED and Photodetector Arrays for AI Optical Links
Avicena, a pioneer in next-generation microLED interconnect technology, has announced a breakthrough at SuperCompute 2025 (SC25). The company revealed that its LightBundle microLED-based links now operate at 4Gbps per lane while using transmitter currents as low as 100µA per LED. At this remarkably low current, each link achieves a raw BER of 1×10⁻¹² and delivers an equivalent energy consumption of just 80fJ/bit per LED, all without requiring Forward Error Correction (FEC). This advancement further reinforces LightBundle as one of the world’s most energy-efficient optical interconnect platforms designed for next-generation AI infrastructure.
Moreover, this milestone directly addresses the growing demands for greater bandwidth, longer reach and increased reliability core requirements as AI workloads and data-intensive applications continue to scale rapidly.
Also Read: ITTech Pulse Exclusive Interview with Sergio Gago CTO at Cloudera
Unlike traditional laser-based optical engines which cannot operate below a minimum lasing threshold microLEDs offer far more flexibility. They can scale their transmit power down to extremely low levels, constrained mainly by the receiver’s signal-to-noise ratio. Additionally, while silicon photonics often relies on splitting a single laser source across several modulators to reduce total power, microLEDs eliminate that complexity by generating their own light. As a result, packaging becomes significantly simpler and more efficient. Avicena’s microLED emitters, only a few microns wide, require no thermal stabilization and avoid complicated control systems. Because they can be densely arrayed, they enable terabit-scale aggregate bandwidth in compact footprints.
Crucially, Avicena’s latest improvement was made possible through advancements in its high-sensitivity receiver technology. Developed in collaboration with manufacturing partners, the new receivers integrate optimized photodetectors adapted from high-volume image sensor processes, ultimately pushing the overall system to achieve unprecedented energy efficiency.
Furthermore, LightBundle takes a different architectural approach by transmitting raw parallel data directly, instead of serializing low-speed on-chip data into ultra-high-speed lanes. This method greatly simplifies IC I/O design and minimizes both latency and power consumption. LightBundle chiplet transceivers also support a wide range of packaging formats, including co-packaged optics (CPO), on-board optics (OBO), wide memory interconnects, and even pluggable modules.
Also Read: IT Tech Pulse Interview Questions for Dr. Zohar Bronfman, CEO & Co-Founder, Pecan AI
Rob Kalman, Chief Scientist at Avicena, emphasized the significance of this achievement: “We already demonstrated an efficient microLED link in a live demo at ECOC 2025 in September. By further optimizing our highly sensitive receivers, we have managed to further reduce the operating currents of the microLEDs and obtained Tx energy consumption down further to tens of femtojoules for this part of the link. Combined with the unique properties of microLEDs, we can achieve unmatched energy efficiency in our LightBundle interconnects. This benchmark shows the scalability of our roadmap, how microLED technology can replace legacy laser-based links with a simpler, more reliable and far lower power solution.”
Ultimately, this development strengthens Avicena’s ongoing collaborations with hyperscale data centers, AI accelerator manufacturers, and memory partners. Together, they are working to enable next-generation AI clusters capable of spanning thousands of GPUs across multiple racks significantly boosting compute performance and supporting the rise of emerging agentic AI workloads.
Write to us [wasim.a@demandmediaagency.com] to learn more about our exclusive editorial packages and programmes.
