DPU / SmartNIC
The smart network card at the edge of every AI server that plugs it into the fabric and handles networking, storage and security so the main chips can focus on compute.
Ranks moderate (57/100) — set apart by cross-layer reach (74) and structural importance (72).
Why it matters
It is the fabric endpoint on every node — the layer that turns nominal link speed into usable, isolated, congestion-managed bandwidth.
Why now
GPU fabrics now ride 400/800G RDMA NICs and DPUs whose generation (ConnectX-7/8, BlueField-3) paces effective bandwidth, congestion behavior and in-network collectives, and sovereign/multi-tenant AI clouds increasingly require the DPU's hardware isolation.
If DPU / SmartNIC runs short
NIC/DPU supply, firmware or congestion-control immaturity throttles effective fabric throughput and utilization even when switches and optics are available.
In depth · editorial + model
Programmable network adapters (SmartNICs) and data processing units (DPUs) that terminate each node's high-speed link and offload RDMA/RoCE, congestion control, storage and tenant isolation from the host CPU. It is the fabric endpoint on every node — the layer that turns nominal link speed into usable, isolated, congestion-managed bandwidth. GPU fabrics now ride 400/800G RDMA NICs and DPUs whose generation (ConnectX-7/8, BlueField-3) paces effective bandwidth, congestion behavior and in-network collectives, and sovereign/multi-tenant AI clouds increasingly require the DPU's hardware isolation. NIC/DPU supply, firmware or congestion-control immaturity throttles effective fabric throughput and utilization even when switches and optics are available.
Who makes DPU / SmartNIC
The companies exposed to DPU / SmartNIC
How to think about it
- The endpoint governs how much of the pipe you actually use
- Offload turns host CPU cycles into network capacity
- Isolation is a hardware feature, not just software
What to watch
- NIC generation shipping with GPU systems (400G→800G)
- In-network compute / collective offload (SHARP) adoption
- DPU-based tenant isolation in sovereign and multi-tenant clouds
- Supply concentration and merchant vs. captive (Nitro/IPU) split
Frequently asked
What is DPU / SmartNIC?
The smart network card at the edge of every AI server that plugs it into the fabric and handles networking, storage and security so the main chips can focus on compute.
Why does DPU / SmartNIC matter for AI?
It is the fabric endpoint on every node — the layer that turns nominal link speed into usable, isolated, congestion-managed bandwidth.
Who makes DPU / SmartNIC?
The companies the model tags as producers or suppliers of DPU / SmartNIC: NVIDIA, Marvell, AMD.
Which companies are most exposed to DPU / SmartNIC?
NVIDIA, Marvell, AMD — 3 companies in total are mapped to DPU / SmartNIC.
What happens if DPU / SmartNIC runs short?
NIC/DPU supply, firmware or congestion-control immaturity throttles effective fabric throughput and utilization even when switches and optics are available.
Where does DPU / SmartNIC sit in the AI value chain?
DPU / SmartNIC sits in the Infrastructure layer of the AI value chain.
Go deeper on DPU / SmartNIC
- The materials, geographies and policies it depends on — heat-mapped
- Substitutes, relief valves and the domino chains if it tightens
- The live tension score, momentum and news drivers
- Four levels of analysis — from plain-English to strategic
model v0.7.0 · research, not advice
Model scores are illustrative reads from our model of the AI value chain — not investment advice.