Topic
AI Networking
Fat-tree, rail-optimized, dragonfly+: which topology survives at 100k GPUs?
- Market Size
- $3.4B
- CAGR 45.0% (3y)
- Companies Tracked
- 6
- Business Signals (90d)
- 0
- Avg Fit Score
- 76/100
This Week
Lead story
Cisco's $1.3B single-quarter webscale AI orders + Silicon One P200 / 8223 launch — the cross-datacenter 'scale-across' training fabric becomes a distinct product category
AI training runs just outgrew the building they live in. The network noticed first.
Read full strategy →Modules
Top Companies by Fit Score
Recent Hiring (top specialties, 90d)
- signal integrity5 roles
- RoCEv23 roles
- network automation2 roles
- co-packaged optics2 roles
- SerDes2 roles
Jobseeker Personas
NETWORK-FABRIC-ARCH-01
AI fabric architect designing rail / fat-tree / dragonfly topologies at 100k+ GPU scale.
NETWORK-SWITCH-ENG-01
Switching silicon engineer at chip vendor or hyperscaler making whitebox builds.
NETWORK-OPS-SRE-01
Site reliability engineer running AI cluster networks day-to-day.
NETWORK-VENDOR-PM-01
Vendor product manager at switching / NIC company prioritising AI features.
Tools
Subscribe to weekly updates on AI Networking
One email Friday morning. Engineers only.
Subscribe to aiinframap Weekly →