AI Server System BOM & Rack Cost — Historical Time Series Data
Full system-level bill of materials for AI training servers — from DGX H100 (8-GPU) through GB200 NVL72 (72-GPU rack). Tracks how rack economics evolved as HBM and networking costs exploded.
AI Server System BOM & Rack Cost — latest readings
8 data points across 4 series, covering DGX H100 (2023) to GB200 NVL72 (2024). Values below are free; the full point-level history is available with Pro.
| Series | Period | Latest ($K) | Change |
|---|---|---|---|
| GPU (COGS) | GB200 NVL72 (2024) | 460.8 | +1632.3% versus DGX H100 (2023) |
| Networking | GB200 NVL72 (2024) | 80 | +900.0% versus DGX H100 (2023) |
| Power & Cooling | GB200 NVL72 (2024) | 120 | +2900.0% versus DGX H100 (2023) |
| Memory, Storage & Chassis | GB200 NVL72 (2024) | 65 | +915.6% versus DGX H100 (2023) |
AI Server System BOM & Rack Cost
Full system-level bill of materials for AI training servers — from DGX H100 (8-GPU) through GB200 NVL72 (72-GPU rack). Tracks how rack economics evolved as HBM and networking costs exploded.
Data Table
| Period | Series | Value |
|---|---|---|
DGX H100 (2023) | GPU (COGS) | 26.6 |
DGX H100 (2023) | Networking | 8 |
DGX H100 (2023) | Power & Cooling | 4 |
DGX H100 (2023) | Memory, Storage & Chassis | 6.4 |
GB200 NVL72 (2024) | GPU (COGS) | 460.8 |
GB200 NVL72 (2024) | Networking | 80 |
GB200 NVL72 (2024) | Power & Cooling | 120 |
GB200 NVL72 (2024) | Memory, Storage & Chassis | 65 |
Methodology & Sources
Synthesized from Epoch.ai hardware cost database, SemiAnalysis system architecture deep dives, and analyst estimates. All costs are approximate manufacture/procurement costs, not list prices.
Citations:
- Epoch.ai / NVIDIA DGX spec (Dec 2023)
- SemiAnalysis (Dec 2023)
- Data center analyst estimates (Dec 2023)
- Silicon Analysts estimate (Mar 2026)
- Silicon Analysts / chipSpecs.ts (Mar 2026)
- SemiAnalysis rack cost analysis (Dec 2024)
- NVIDIA NVL72 spec / Silicon Analysts (Mar 2026)
Related Tools
Related Analysis
Market intelligence on this data
All market intelligence →TSMC CoWoS Yield Confirmed at 98-99% in HVM; Nvidia Weighs 8-Stack HBM4E Fallback for Rubin Ultra
CoWoS yield: 98-99% confirmed in HVM (positive); HBM4E 12-stack yield risk flagged — potential stack-count reduction from 12 to 8 layers (est. HBM capacity-per-die reduction of ~33% if downgrade confirmed); TSMC 2026 CapEx raised to $60-64B record (capacity positive).
Nvidia Rubin Ultra HBM Spec Downgrade Under Review as HBM4/4E Supply Crunch Deepens Through 2027
Customer HBM4 fill rates at 60–70% of ordered volumes (Micron-confirmed); HBM4 BOM cost for GPU vendors estimated at ~2x HBM3E per Korean press citing industry sources; CoWoS lead times confirmed at 52+ weeks; ABF substrate orders extending into 2028.
Nvidia–SK Hynix Multi-Year HBM Supply Partnership Confirmed at Up to $500B
Supply commitment: up to $500B multi-year; Capacity: SK Hynix HBM allocation heavily directed toward Nvidia; Lead time risk for non-Nvidia HBM buyers: elevated (magnitude unquantified in sources).
Nvidia-SK Hynix & Samsung-Broadcom Lock In $950B HBM Supply Agreements, Securing Next-Gen Memory Pipeline Through 2030
SK Hynix: $750B long-term HBM supply commitment to U.S. customers (led by Nvidia); Samsung: $200B memory + foundry + packaging commitment to Broadcom through 2030; SK Telecom 2GW AI data center powered by HBM4 targeting 2027 online date; TSMC CoWoS and advanced packaging capacity described as largely booked through 2027.