Specifications
Family
Lightmatter Passage
Technology
3D co-packaged optics (CPO)
Passage L200 Total Bandwidth
256 Tbps (Tx+Rx) per package
Passage L200 Target
Next-generation XPUs and switches
Passage L200 Connectors
Detachable fibre connectors
Aggregate Bandwidth
32 to 64 Tbps through co-packaged optics
Signalling
112G PAM4
Passage L20 Type
High-bandwidth bidirectional photonic link reference platform
Passage L20 Rate
Up to 3.2 Tbps per fibre at 1.9 pJ/bit
16-Wavelength Link
800 Gbps per fibre at 2.6 pJ/bit
Passage M1000 EVK
Rack-level 3D photonic interposer reference platform
Passage M1000 EVK Capacity
114 Tbps across a 4,000 mm2 footprint
Other Reference Systems
Passage EVK100 and Passage EVK50
Architecture
Edgeless I/O - removes faceplate bandwidth limits
Target Workloads
Frontier-scale AI training and inference
Vendor
Lightmatter
Overview
Lightmatter Passage is a 3D co-packaged optics family that moves the optical interconnect inside the package instead of behind a faceplate. The Passage L200 delivers 256 Tbps of aggregate Tx+Rx bandwidth per package with detachable fibre connectors, and supports 32 to 64 Tbps of co-packaged bandwidth using 112G PAM4 signalling - designed for next-generation XPUs and switch silicon.
The Passage L20 reference platform pushes per-fibre performance: a high-bandwidth bidirectional link running up to 3.2 Tbps per fibre at 1.9 pJ/bit, and a 16-wavelength bidirectional link delivering 800 Gbps per fibre at 2.6 pJ/bit. For system architects, the Passage M1000 EVK provides a rack-level 3D photonic interposer reference platform reaching 114 Tbps across a 4,000 mm2 footprint, with Passage EVK100 and EVK50 as smaller evaluation platforms.
Together these products address the faceplate bottleneck in AI-scale networks, where aggregate switch bandwidth doubles every generation while discrete pluggable optics consume faceplate area, power and signal-integrity margin. Lightmatter's edgeless I/O approach is intended to scale interconnect bandwidth per package rather than per faceplate slot.
Key Benefits
Faceplate-free scaling. Co-packaged optics remove the bandwidth ceiling imposed by pluggable faceplate links. Record per-fibre efficiency. Up to 3.2 Tbps per fibre at 1.9 pJ/bit on the L20 platform. Rack-level validation. The M1000 EVK demonstrates 114 Tbps across a 4,000 mm2 interposer for architecture evaluation. Serviceable optics. L200 detachable fibre connectors keep the optics maintainable in production systems.
Applications
AI datacenter switch interconnect, scale-up and scale-out XPU fabrics, frontier-scale training clusters, optical interposer research, and next-generation co-packaged optics platforms.
Request a Quote — LIGHTMATTER PASSAGE — 3D CO-PACKAGED PHOTONIC INTERCONNECT FOR XPUS AND SWITCHES
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →