GPU Depreciation Curves & Schedule Calculator

H100, H200, A100, B200 Depreciation Rate Analysis

1What It Measures

Exponential decay curves for datacenter GPU residual value over 5 years. Models H100, H200, A100, B200, MI300X, and legacy GPUs with configurable purchase prices.

2Why It Matters

GPU depreciation directly impacts loan-to-value ratios for GPU-backed financing, lease buyout decisions, and total cost of ownership models. Understanding how fast H100s depreciate is critical for private credit underwriting.

3How to Read It

Select GPUs to compare, adjust purchase price, and view depreciation curves. Half-life indicates months to 50% value. Floor value is the minimum residual (typically 10-15% for scrap/secondary market value).

Key Metrics Explained
Models Compared
3

GPU types selected

Purchase Price
$40K

Base price per unit

Fastest Decay
16mo

Shortest half-life

Select GPUs to Compare

Current Generation

Previous Generation

Legacy

$

GPU Depreciation Curves Comparison

Residual value in USD over 60 months

Unlock Full Depreciation Schedule

Unlock the complete depreciation milestones table, floor values, and multi-year residual analysis for all GPU models.

Model Calculation Disclaimer

These calculations use simplified models of complex market realities. Assumptions about future conditions are inherently uncertain. Small changes in input parameters can significantly affect outputs. Always verify results with qualified professionals.

View methodology

How Fast Do H100 GPUs Depreciate?

GPU depreciation follows an exponential decay curve, not straight-line depreciation. The “half-life” represents the time for a GPU to lose 50% of its value above the floor. Understanding H100 depreciation rates, H200 residual curves, and A100 depreciation schedules is critical for GPU-backed loan underwriting and lease vs. buy decisions.

H100 Depreciation Rate

NVIDIA H100 SXM GPUs have a ~20-month half-life. At $40,000 MSRP, expect ~$24,000 residual at 12 months and ~$17,000 at 24 months. The H100 depreciation curve flattens near the 12% floor value (~$4,800) as secondary market demand for AI training hardware remains strong.

H200 vs H100 Depreciation

The NVIDIA H200 has a slightly longer half-life (22 months) due to HBM3e memory upgrades extending its competitive window. However, B200/Blackwell announcements may accelerate H200 depreciation. Compare H200 depreciation curves against H100 to model upgrade timing and refinancing opportunities.

A100 Residual Value 2026

NVIDIA A100 GPUs launched in 2020 are now 5+ years old. A100 depreciation has largely plateaued near floor values (10-12%). Secondary market A100 prices are $3,000-5,000 for SXM variants, primarily driven by inference workloads that don't require latest-gen silicon.

Frequently Asked Questions

What is the H100 depreciation rate per year?

H100 GPUs depreciate approximately 35-40% in Year 1 and 20-25% in Year 2. This is faster than typical IT equipment (20-25%/year) due to rapid AI hardware iteration cycles. By Year 3, depreciation slows as values approach floor levels.

How do lenders calculate GPU collateral value for GPU-backed loans?

Lenders apply loan-to-value (LTV) ratios of 50-70% to current market value, then require monthly principal payments that match or exceed the depreciation rate. GPU-backed loan terms typically range from 18-36 months with reappraisal triggers at 6-month intervals to maintain target LTV ratios.

When should I exercise a GPU lease buyout?

Exercise a lease buyout when the buyout price is below current market residual value AND you have a use case for the hardware. Compare the buyout price against secondary market prices (CloudHardware, eBay enterprise) and factor in remaining useful life for your workloads. Generally, buyout makes sense when buyout price is <80% of market value.

How does Blackwell B200 impact H100 depreciation?

New GPU generations typically accelerate depreciation of previous generations by 10-20% in the 6 months following launch. B200's 2025 release may compress H100 half-life from 20 to 16-18 months. However, supply constraints for new chips often support residual values of current-gen hardware during the transition period.