Persistent AI demand for high-memory Hopper GPUs like the H200, driven by large language model training and inference workloads, underpins the closely matched leading outcomes around $6.50–$7.00 and $7.50+ per GPU-hour at 18.4% each. Current on-demand medians hover near $4.50–$5.50 across providers, with spot rates often lower, reflecting a year of upward pressure despite recent spot softening. NVIDIA’s shift of TSMC capacity away from H200 production toward the Rubin architecture, alongside limited new Hopper supply as Blackwell ramps, sustains scarcity expectations through year-end. Key near-term catalysts include NVIDIA’s upcoming earnings and any Blackwell utilization data that could clarify whether demand sustains elevated rates or enables supply relief.
基於Polymarket數據的AI實驗性摘要。這不是交易建議,也不影響該市場的結算方式。 · 更新於1H6H1D7D30DALL
View resolved

警惕外部連結哦。
警惕外部連結哦。
Frequently Asked Questions