Inventory Control
Operations
March 26, 2026
Centralized inventory control for GPU infrastructure assets with 24-hour restoration target
288
GPU modules per facility
(4 x GR200 NVL72)
(4 x GR200 NVL72)
24h
Single-failure restoration
target from on-site stock
target from on-site stock
$45M
Total asset value
under insurance coverage
under insurance coverage
0
Spare GPU modules
NVIDIA contract covers
NVIDIA contract covers
Supplier Lead Times
Minimum Stock Thresholds
GPU Modules
0 units
No spares — unit cost ~$90K+
NVIDIA maintenance contract
NVIDIA maintenance contract
IB Transceivers
5%
Of installed base
~15 units reserve
~15 units reserve
NVMe Drives
3%
Of installed capacity
~12–15 hot-swap reserve
~12–15 hot-swap reserve
DLC Components
1 set
Full pump assembly
+ fittings kit on-site
+ fittings kit on-site
UPS Batteries
1 set
Replacement set per
maintenance cycle
maintenance cycle
Service Provisioning Lead Time
T1
Pay-per-Use
< 1 hour
Automated API provisioning from available pool
T2
Reserved Capacity
1 – 3 days
Partition allocation + network isolation + credentials
T3
Private AI Factory
0 days
Pre-provisioned — client onboarding = access config
GF
Greenfield Facility
12 – 18 months
Full build cycle — investment-level, anchor LOI required
Information Flow & Systems
Sales / BD
→
Head of Infra
→
Remote NOC
→
Procurement
Grafana + Prometheus
GPU / thermal / network telemetry
DCIM
Physical asset tracking & serial inventory
Cloud ERP
Procurement & financial records
NVIDIA Contract
Annual maintenance subscription — $400K/yr
Loss Prevention & Security
Physical Controls
- Remote mountain location
- Perimeter fencing + biometric access
- 24/7 CCTV — NOC-monitored
- Access log — all visits authorized
- No permanent on-site staff
Digital Controls
- GPU firmware / BIOS locked
- 802.1X + VLAN isolation
- Telemetry detects tampering
- All assets serialized in DCIM
- Client workload isolation
Audit & Insurance
- Quarterly physical audit
- $45M infrastructure insurance
- NVIDIA maintenance contract
- Reduces resale incentive
