Table of Contents
Unveiling the Surging Residual Value of NVIDIA’s GPUs in the AI Super Cycle
The age of LLMs has ushered in a new era where NVIDIA’s GPUs serve as a key indicator of the AI super cycle’s vitality. Recent data from Silicon Data reveals a robust upcycle, defying calls for a slower pace in AI model development from industry experts like Anthropic’s Dario Amodei and OpenAI’s Sam Altman.
NVIDIA’s B200 GPUs Residual Value Surpasses 158% of Launch Price
Silicon Data’s analysis reveals that NVIDIA’s older A100 and H100 GPUs are commanding a residual value well above traditional depreciation schedules. This value exceeds the predicted depreciation based on an estimated useful life. NVIDIA’s B200 GPUs, in particular, are currently trading at a remarkable 158% premium over their original launch price, showcasing a strong demand and scarcity in the market.
One of the driving factors behind this high residual value is the B200’s efficiency in handling inference workloads. Estimates from SemiAnalysis indicate that the GPU costs only $0.20 per 1 million tokens at 76 tokens per second per user on platforms like DeepSeek R1, making it a cost-effective solution for AI tasks.
Evolution of GPU Rental Prices
In August, Silicon Data’s GPU rental price data illustrated a significant surge in rates for NVIDIA’s B200 GPUs. From hovering just below $5 per hour in January, the hourly rates had escalated to between $5.50 and $5.80 by August, representing a notable appreciation of 50% to 80%.
About the author: With over 2,200 articles on financial and tech topics, spanning nearly 1 million words, I have been a member of the WccTech mobile team since 2025. As a graduate of the University of Toronto’s Rotman Commerce Program, I offer nuanced insights and a unique perspective on various subjects. When not writing, I indulge in my passion for exploring hidden confectionaries and restaurants as an aspiring food connoisseur.
Follow Wccftech on Google for more news coverage.
Conclusion: The surging residual values of NVIDIA’s GPUs, particularly the B200 series, underscore the growing demand and efficiency of these devices in the AI infrastructure landscape. As the industry continues to evolve, these trends offer valuable insights into the market dynamics and the evolving needs of AI startups and compute providers.