GPU 能保值嗎？

GPU 持續賺錢的時間，遠比它維持每小時費率的時間長。2020 年推出的卡到 2029 年仍有合約收入，NVIDIA 也表示這些卡如今仍滿載運行；但一小時的租用價格波動劇烈，從 2023 年的短缺高峰下跌約 80%，之後才回升。壽命長，價格波動大。

一小時 H100 的價格變化

一款 2020 年推出的 GPU，合約收入已延續到 2029 年，也就是九年的賺錢壽命。: “CoreWeave signed a contract for Nvidia A100 GPUs that runs into 2029, nine years after the Ampere SKU debuted.” (Mike Intrator，CoreWeave 執行長，於第二季財報電話會議)
NVIDIA 表示，六年前的晶片並沒有閒置，而是滿載運行。: “Thanks to CUDA, the A100 GPUs we shipped six years ago are still running at full utilization today, powered by a vastly improved software stack.” (Colette Kress，NVIDIA 財務長)
合約到期的硬體重新出租時，價格接近原始價格，而不是打折出租。: “A batch of H100s coming off an expired contract was immediately rebooked at 95% of the original price.” (Mike Intrator，CoreWeave 執行長，接受 CNBC 訪問)
決定退役日期的是軟體堆疊，而不是晶片本身：CUDA 相容性讓舊卡得以留在機群中。: “CUDA's compatibility and our massive installed base extend the life in our systems well beyond their original estimated useful life.” (Colette Kress，NVIDIA 財務長)

一位知名放空者認為，六年的折舊時程是會計操作而非工程現實，業界也因此高估了獲利。: “Understating depreciation by extending useful life of assets artificially boosts earnings.” (Michael Burry，他估計 2026 至 2028 年間此效應在各超大規模雲端業者合計達 $176bn)
有一家超大規模雲端業者反其道而行：Amazon 將部分伺服器的耐用年限從六年「縮短」為五年，理由是 AI 硬體的發展速度。: “An increased pace of technology development, particularly in the area of artificial intelligence and machine learning.” (Amazon，於 2 月的一份申報文件中)
當代顯示卡的租賃費率可能跌得又快又猛。隨著供給追上，H100 每小時價格從 2023 年的短缺高峰崩跌。: “Rental costs dropped from $8 per hour in 2023 to $1–$2 in 2024 and 2025 due to oversupply.”
NVIDIA 大約每年推出一代新架構，比上述任何折舊時程都快。