DDNDDN
Opportunities

Pacific Compute Services

ValidateActiveInfiniaHorizonAPACCloud
Inference response times & tenant fairness
$800.0K
80
Qualification
0
Solution Fit
0
Discovery
0
Value Conf.
90
Tech Confidence
Industry
Telecommunications
Country / Region
Singapore / APAC
Close Date
—
Sales Owner
—
SE Owner
—
Probability (SE judgement)
55%
Quote Status
—
Contract Value
$2.40M
Next Action
Complete tenant fairness benchmark
Owner: James Tan · Due: 2026-12-08
Business Challenge & Consequences

Inference response times and tenant fairness under load.

SLA breaches; customer churn risk.

Desired Business Outcomes

Consistent inference latency; tenant fairness; lower cost per token.

Compelling Event

New enterprise tenant onboarding

Workloads & Solution Direction
Inference and long-context/agentic workloadsAI-as-a-service and multi-tenancy
Primary Direction
multi-tenant-ops
Application / Framework
—
Current Environment
Compute / GPU
Shared H100 pool
GPU / Node Counts
96× H100
Storage / Data Platform
Generic object store
Network
400Gb Ethernet
Workload Characteristics
Data Capacity & Growth
—
Active Working Set
70 TB
Current Performance
TTFT 800ms p95, fairness variance 25%
Target Performance
TTFT 350ms p95, fairness variance < 10%
Access Patterns
Object, high-concurrency
Read/Write, Concurrency
—
Required Protocols
S3, gRPC
Checkpoint / Restart
—
Availability / Recovery
—
Security / Residency
—
Known Technical Blockers
None blocking
Success Criteria (agreed)
p95 TTFT < 400ms; fairness variance < 12%
Competition & Alternatives
Competitors
Hyperscaler inference
Alternatives
Add more GPUs