Jensen Huang compares old graphics card GPUs to fine wine "improving with age": five years ago, chip values continued to rise, increasing faster than aged wine

The explosive demand for AI has driven GPU prices to rise across the board, with even chips from five years ago continuing to appreciate. NVIDIA CEO Jensen Huang describes this phenomenon as the “Fine Wine Effect,” meaning that old GPUs are appreciating in value even faster than aged fine wine.
(Background summary: GPU computing power shortage repeats: major companies like OpenAI and Anthropic consume supply, AI startups are queued until the end of the year)
(Additional background: Is the GPU bubble about to burst? Experts say: NVIDIA H100 leasing prices have collapsed to “$2 per hour,” why has demand sharply decreased?)
As AI demand enters a period of explosive growth, GPUs have become one of the most scarce computing resources worldwide, even older chips circulating for 4 to 5 years are beginning to see continuous price increases. NVIDIA CEO Jensen Huang describes this phenomenon as the “Fine Wine Effect,” implying that old GPUs are like aged fine wine—becoming more valuable and more fragrant over time.

The Origin and Evolution of the “Fine Wine Effect”

According to foreign media reports, the term “Fine Wine” first appeared during the era of AMD Vega graphics cards, used to describe how GPU performance improved through continuous driver optimization. However, with the advent of the AI era, the meaning of the term has fundamentally changed—now it is more often used to describe the phenomenon of GPU prices rising over time.
Huang pointed out that the price increase rate of old GPUs even surpasses the aging speed of real fine wine, and there remains strong demand in the market. Even products from 4 to 5 years ago are still appreciating in value.

AI Computing Power Demand Sparks GPU Supply and Demand Imbalance

NVIDIA stated that AI demand is driving the GPU market into a period of explosive growth. Almost all data centers engaged in AI operations worldwide rely on GPUs as their core computing resource. Although CPU demand is also growing, GPUs still hold an irreplaceable position.
Currently, global semiconductor capacity is tight across the board—from wafer manufacturing to GPUs, CPUs, DRAM, and AI cloud computing platforms, the entire industry chain faces supply shortages. This tense situation has further driven up the prices of various hardware, including GPUs, with older models also affected.

CoreWeave CEO: Demand for Old GPUs Accelerates Growth

Mike Intrator, CEO of cloud computing giant CoreWeave, echoed Huang’s view. He pointed out that demand for old GPUs is accelerating, with prices for models like H100, H200, L40S, and A100 rising compared to the previous quarter. Intrator said that CoreWeave’s computing resources have been sold out long-term, and this situation is common across the industry.
Analysts believe that as the number of AI model parameters continues to grow and enterprise AI deployment accelerates, the supply-demand imbalance for GPUs will be difficult to resolve in the short term. Even “outdated” old GPUs remain hot in the second-hand and cloud rental markets, giving real-world support to Huang’s “Fine Wine Effect” analogy.