What is a scalable unit in an AI factory?
Why C is the answer
A repeatable block of compute and network
Reference designs scale by adding units.
In detail
- Reference designs scale by adding units.
- Reference designs scale this way.
Use case
An illustrative scenario, not a real incident.
An AI factory grows by adding scalable units, each a known block of compute and network.
See it in different ways
flowchart TD L1["Leaf for rail 1"] --- A["Server A NIC 1"] L1 --- B["Server B NIC 1"] L2["Leaf for rail 2"] --- C["Server A NIC 2"] L2 --- D["Server B NIC 2"]
mindmap
root((AI factory))
Scale up
NVLink
Scale out
Rail-optimized fabric
Building block
Scalable unit
Management
Out of bandmindmap
root(("AI factory"))
n1["Scalable unit 1"]
n2["Compute"]
n3["Network"]
n4["Scalable unit 2"]
n5["Compute"]
n6["Network"]Key point: Scalable units repeat.
Why the other options are wrong
- A. A software license for one user
- It is not a license.
- B. A kind of optical transceiver
- It is not a transceiver.
- D. A single GPU with its own power supply
- It is much larger than one GPU.