Scintil Releases DWDM Laser Source Evaluation Kit for Scale-Up AI Networks

Why It Matters

By delivering half the power and reduced latency needed for massive AI processor clusters, the LEAF Light EVK accelerates the shift from copper to optical interconnects, a critical bottleneck for hyperscale AI deployments. The funding and foundry partnership position Scintil for high‑volume production, influencing the future of AI data‑center architecture.

Key Takeaways

• LEAF Light cuts power 50% vs single-wavelength CPO.
• Evaluation kit offers 8‑fiber, 8/16‑wavelength DWDM testing.
• On‑chip WaveGuard ensures precise wavelength spacing.
• Series B $58M funding includes NVIDIA, signaling market confidence.
• SHIP platform validated on Tower’s 200 mm silicon photonics line.

Pulse Analysis

The rapid expansion of generative‑AI models has pushed data‑center interconnects to the brink of copper’s physical limits. While single‑rack systems still rely on electrical links, multi‑rack and hyperscale clusters demand far higher bandwidth density, lower power consumption, and deterministic latency. Dense wavelength‑division multiplexing (DWDM) co‑packaged optics (CPO) has emerged as the most promising solution because it can multiplex dozens of channels over a single fiber, dramatically increasing throughput without adding cabling complexity. However, the technology has remained confined to laboratory prototypes, leaving operators uncertain about real‑world integration and scalability.

Scintil’s LEAF Light chip addresses those concerns by integrating a full DWDM array on a single silicon die, delivering up to 50 % power savings versus conventional single‑wavelength CPO and eliminating the need for heavy forward error correction. The accompanying Evaluation Kit (EVK) gives customers an end‑to‑end testbed with eight fibers, configurable 8‑ or 16‑channel LOSAs, and on‑chip telemetry such as WaveGuard frequency trimming and per‑wavelength power control. By automating wavelength lock and power uniformity, the kit shortens validation cycles and provides the latency‑critical performance data that system architects require before committing to volume production.

The $58 million Series B round, anchored by NVIDIA, signals strong market confidence that DWDM‑based optical fabrics will become the backbone of future AI infrastructure. Scintil’s partnership with Tower Semiconductor, which has already qualified the SHIP manufacturing process on 200 mm silicon‑photonic lines, paves the way for high‑volume, cost‑effective deployment. As hyperscale cloud providers and AI‑focused enterprises evaluate the EVK, the competitive landscape is likely to shift, pressuring incumbent optical vendors to accelerate their own integrated‑photonic roadmaps. Successful adoption could redefine data‑center economics, lowering power bills while unlocking the bandwidth needed for next‑generation AI workloads.