Blackwell GPU rents spike

Hourly rental rates for Nvidia Blackwell GPUs have jumped sharply, signalling tight supply at the high end of AI compute markets. An index showed rates hit $4.08 per hour, up 48% from $2.75 two months earlier and linked the rise to demand for agentic AI workloads. Reports say the price movement is a live market signal of scarcity for premium inference capacity. (alltoc.com) (intellectia.ai)

Renting Nvidia’s newest Blackwell graphics processors now costs about $4.08 an hour, up 48% from $2.75 two months ago. (techmeme.com) The price jump comes from the Ornn Compute Price Index, which tracks spot rental prices for Nvidia chips in cloud data centers. Other reports on Monday said prices across Nvidia’s wider graphics processor lineup have also risen in recent months. (techmeme.com) (fxbus.com) A spot rental market is the short-notice market for computing power, where companies pay by the hour instead of signing long contracts. When hourly prices rise that fast, it usually means customers want immediate access to scarce machines. (fxbus.com) Blackwell is Nvidia’s latest family of artificial intelligence chips, built for training models and running inference, the stage where a model answers prompts after it has been trained. Amazon Web Services said in May 2025 that its Blackwell-based P6-B200 instances offered up to twice the performance of its earlier P5en systems for machine learning and high-performance computing. (aws.amazon.com) Cloud providers spent much of 2025 rolling Blackwell into commercial service. CoreWeave said on February 4, 2025 that it was the first cloud provider to make Nvidia GB200 NVL72 instances generally available, and Microsoft later announced general availability for its ND GB200 v6 virtual machines. (coreweave.com) (techcommunity.microsoft.com) Google and Amazon both tied Blackwell launches to reasoning models, which generate extra tokens while working through a problem, and to agentic systems, which chain together multiple model calls to complete tasks. Those workloads consume more memory, networking, and compute time than a simple chatbot reply. (cloud.google.com) (aws.amazon.com) Nvidia and Oracle also said in late 2025 that thousands of Blackwell graphics processors were being deployed for reasoning models and artificial intelligence agents on Oracle Cloud Infrastructure and Nvidia DGX Cloud. That expansion did not stop prices from rising in the hourly market. (blogs.nvidia.com) The current signal from the rental market is simple: even after major cloud launches, premium Blackwell capacity is still expensive to get on demand. As long as companies keep chasing low-latency inference for reasoning and agentic tools, hourly prices will keep showing where supply is tightest. (techmeme.com) (aws.amazon.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.