GPU Infrastructure // Qualification Portal
Client Qualification Brief

Large-Scale GPU
Cluster Request

Complete this brief prior to pricing, allocation, or delivery discussions. All fields are classified as Mandatory (M) or Desirable (D) to ensure rapid alignment with current inventory including B200/B300 capacity.

Active: H100 256-node / InfiniBand
Active: H200 256-node / InfiniBand
Active: A100 256-node
Available: B200 / B300 Capacity
Classification: M Mandatory requirement D Desirable / flexible Toggle each field's classification as needed

0
Contact Information
Who should we follow up with?
Full Name *
Company / Organization *
Email Address *
Phone / Signal
Role / Title
Priority Qualification Questions
Complete these four questions first — required for rapid first-pass assessment
RAPID FIRST PASS
Is H100, H200, or A100 an exact SKU requirement — or are equivalent / newer GPUs (B200, B300) acceptable?
✓ Exact SKU only
✓ Newer equivalents acceptable
✓ Open to any equivalent compute
✓ Prefer B200/B300
Is InfiniBand mandatory — or would high-speed Ethernet (400G RoCE) be acceptable?
✓ InfiniBand mandatory
✓ High-speed Ethernet acceptable
✓ Either acceptable
✓ No preference
Is the 256-node quantity fixed — or is equivalent total GPU count or aggregate compute capacity acceptable?
✓ Fixed at 256 nodes
✓ Equivalent GPU count acceptable
✓ Total FLOPS / compute matters more
✓ Flexible
Are there location or delivery timeline constraints?
Preferred region / geography
Required delivery / start date
3.1
GPU Requirements
Model, quantity, form factor, and compute specifications
M
✓ A100
✓ H100
✓ H200
✓ B200
✓ B300
✓ Open / equivalent
M
✓ Exact SKU required
✓ Equivalent acceptable
✓ Newer generation preferred
M
Is this fixed, or is flexibility permitted based on total GPU count?
D
✓ SXM
✓ PCIe
✓ No preference
M
M
✓ Bare metal
✓ MIG
✓ Virtual GPU
✓ No requirement
3.2
Networking Requirements
InfiniBand, topology, RDMA, and bandwidth specifications
M
✓ Mandatory
✓ Preferred
✓ Optional
✓ High-speed Ethernet acceptable
D
D
✓ Non-blocking required
✓ Blocking acceptable
✓ No requirement
D
✓ RDMA required
✓ NCCL optimized
✓ High East-West BW
✓ None specified
3.3
CPU & Memory Requirements
Host processor, RAM, and interconnect specifications per node
D
✓ Intel (Sapphire Rapids+)
✓ AMD (Genoa/Turin+)
✓ ARM (Grace)
✓ No preference
D
M
D
✓ Yes
✓ No
✓ Unknown
D
3.4
Storage Requirements
Local NVMe, shared storage, throughput, and checkpointing
D
✓ Required
✓ Preferred
✓ Not required
D
D
✓ Lustre
✓ BeeGFS
✓ NFS
✓ Object storage
✓ None required
D
D
3.5
Deployment & Location
Geography, timeline, commitment term, and delivery requirements
D
✓ US East
✓ US West
✓ US Central
✓ EU
✓ APAC
✓ Flexible
M
✓ Single-site required
✓ Multi-site acceptable
✓ No preference
M
✓ Cloud / hosted
✓ Bare metal
✓ Dedicated cluster
✓ On-premises
M
M
✓ Monthly
✓ 6 months
✓ 1 year
✓ 2 years
✓ 3 years
✓ 5+ years
✓ TBD
3.6
Commercial & Compliance Constraints
Budget, compliance, OEM preferences, and infrastructure constraints
M
M
✓ Dedicated / no sharing
✓ Multitenant acceptable
✓ No restriction
D
✓ SOC 2 Type II
✓ ISO 27001
✓ FedRAMP
✓ HIPAA
✓ ITAR / EAR
✓ GDPR / EU residency
✓ None
D
D
D
Confidential. Information provided is used solely for infrastructure qualification purposes. SigmaEdge does not share client requirements with third parties without explicit authorization.
✓
Brief Received
Your qualification brief has been submitted. A member of the SigmaEdge infrastructure team will review your requirements and be in touch within one business day.