Case Study

65 Nodes. Under Four Weeks. Cloud to On-Prem, No Downtime

How Arc Compute architected, procured, and delivered a 65-node NVIDIA HGX H100 cluster for Boson AI, beating a 12-week industry lead time by two-thirds.

Client
Boson AI
Industry
Generative AI
Use Case
LLM Training and Inference for AI Voice Agents
Inside the case study

What you'll find in the Case Study

A case study showing how Boson AI migrated from cloud GPUs to a 65-node NVIDIA HGX H100 cluster in under four weeks, accelerating LLM development while reducing infrastructure costs.

65x
NVIDIA HGX H100 Nodes
4Wk
Delivery Time
520
CX-7 NICs Installed
01

The cloud economics that drove Boson’s move to on-prem infrastructure.

02

How to deploy a production-scale H100 cluster during a global GPU shortage.

03

The procurement strategy that beat industry lead times by more than two-thirds.

04

The networking and cluster architecture powering large-scale AI training and inference.

05

How Arc overcame a critical 520-NIC supply chain challenge mid-deployment.

06

Four infrastructure lessons for organizations building AI clusters at scale.

Arc Compute helped us stand up our infrastructure faster than we thought possible, and support has been outstanding from day one.

Access the case study

Fill out the form below and we'll send the case study PDF directly to your inbox.

Thank you! Your submission is being processed and you will receive the case study shortly.
Oops! Something went wrong while submitting the form.