Solutions
Optical interconnect for AI infrastructure
From GPU cluster fabrics to data center interconnect, we supply the 400G / 800G / 1.6T optics that build modern AI compute — in volume, with project support.
AI / GPU cluster fabric
Spine-leaf and rail-optimized fabrics for GPU training clusters using 400G and 800G DR/SR optics with breakout.
Data center interconnect
FR4 / LR4 optics for campus and metro DCI links up to 10 km, connecting availability zones and pods.
Low-power & LPO
Silicon-photonics and linear-drive (LPO) options to cut power and latency in dense high-speed fabrics.
Reference architecture
Spine-leaf fabric for AI clusters
A typical AI cluster uses 800G optics at the spine and 400G at the leaf, with 800G→2×400G or 400G→4×100G breakout to servers and NICs. PhoScale supplies every layer of this fabric from a production-ready product line.
- Spine / super-spine: 800G OSFP DR8, 1.6T OSFP DR8
- Leaf: 400G / 800G DR4 / DR8, breakout-ready
- Server / NIC: 400G QSFP-DD DR4, SR4/VR4 short reach
- DCI: 400G FR4 / LR4 for 2–10 km links
Project support
Built for project-based procurement
BOM
review & quoting
Phased
batch delivery
Coding
compatible on request
Alloc.
capacity locking