Drut Technologies has announced a strategic collaboration with Ranovus to deliver a commercially available co-packaged optics (CPO) solution aimed at next-generation AI, ML, and high-performance computing (HPC) workloads. The integration of Ranovus’s ODIN CPO platform into Drut’s 2500 series architecture enables direct PCIe Gen 5 server-to-photonic fabric connectivity at full x16 line speed (32G per serdes), significantly improving bandwidth density and reducing power consumption in large-scale compute clusters.
DRUT Technologies, headquartered in Ottawa, Canada, is an infrastructure technology company focused on building dynamic, software-defined photonic fabrics for data center environments optimized for AI, machine learning (ML), and high-performance computing (HPC). The company’s architecture centers on an all-photonic infrastructure that replaces traditional electrical interconnects, aiming to reduce latency and energy consumption while improving bandwidth scalability. DRUT’s flagship offering, the 2500 Series, integrates PCIe Gen 5 server connectivity with optical switching to enable composable infrastructure at scale.
Ranovus’s ODIN solution consolidates key optical engine components—including quantum dot lasers, silicon photonics, and mixed-signal ASICs—into a single monolithic EPIC (Electro-Photonic Integrated Circuit). This scalable technology supports CPO, Near-Packaged Optics, and pluggable module use cases, addressing the growing demand for energy-efficient and high-bandwidth interconnects in hyperscale environments.
The partnership signals growing momentum for CPO adoption in PCIe Gen 5/6/7 and 100Gbps/200Gbps Ethernet-based AI clusters. By tightly coupling Drut’s control-plane architecture with Ranovus’s optical innovation, the solution provides a path to sustainable, high-performance infrastructure capable of supporting increasingly data-intensive applications.
- Drut’s 2500 series now supports full x16 PCIe Gen 5 over photonic fabrics
- Ranovus ODIN® integrates lasers, modulators, detectors, and ASICs into compact EPIC
- Solution addresses power, latency, and footprint challenges of traditional interconnects
- Designed for hyperscale and AI compute clusters using 100G/200G Ethernet
- Supports roadmap to Gen 6/7 PCIe and future AI workloads at scale