Table of Contents
The Rise of NVIDIA GPU Rental Rates: A Closer Look at the AI Build Out
The world of artificial intelligence (AI) is constantly evolving, with the appropriate depreciation rate for NVIDIA GPUs being a key factor in this ongoing transformation. The debate between proponents advocating for a longer shelf life and detractors pushing for a 3-year useful life estimate has been at the forefront of discussions. However, recent data on hourly rental rates for NVIDIA GPUs suggests a shift in perspective.
Older NVIDIA GPUs Commanding Higher Rental Rates
Surprisingly, older NVIDIA GPUs such as the A100s and H100s are now demanding a higher rental rate compared to the beginning of the year. Silicon Data reports that NVIDIA H100 GPUs, which were renting for just under $2 per hour in January 2026, have now surpassed the $2 mark and are nearing $3. Similarly, NVIDIA’s B200 GPUs, previously priced below $5 per hour, are now ranging between $5.50 and $5.80.
Efficiency Gains and Value Retention
The driving force behind this unexpected price surge lies in the increasing efficiency of AI models running on existing compute infrastructure. This efficiency allows older NVIDIA GPUs to retain their value, following a Jevons paradox-type paradigm where technology adoption accelerates as barriers diminish.
Implications for Hyperscalers
These market dynamics have significant implications for hyperscalers like Microsoft, enabling them to extend the useful life estimate of their NVIDIA GPU inventory. By reducing annual GPU depreciation expenses, hyperscalers can bolster their net income, contributing to a positive financial outlook.
As the AI landscape continues to evolve, the value and longevity of NVIDIA GPUs remain pivotal in shaping the industry’s future.