Blackwell GPU hours jump 48%

Hourly rental rates for Nvidia’s newest Blackwell GPUs have climbed to $4.08 from $2.75 two months ago — a roughly 48% increase as demand for agentic AI grows (intellectia.ai). The rise signals tighter supply and higher access costs for teams needing premium AI compute (intellectia.ai).

Renting Nvidia’s newest Blackwell graphics processors now costs about $4.08 an hour, up from $2.75 two months ago, according to the Ornn Compute Price Index. (techmeme.com) The move amounts to a roughly 48% jump in spot rental prices for Blackwell chips in cloud data centers as of April 13, 2026. The Wall Street Journal report, cited by Techmeme and other outlets, tied the increase to heavier demand for “agentic” artificial intelligence systems that run more steps per task. (techmeme.com) (the-decoder.com) A graphics processor, or GPU, is the specialized chip that trains and runs large artificial intelligence models. Nvidia’s Blackwell line is its latest data-center generation, sold in products such as the B200 and GB200 systems now offered by cloud providers including CoreWeave, Lambda, Oracle Cloud Infrastructure, and Nvidia’s own cloud partners. (nvidia.com) (lambda.ai) (nvidia.com) Blackwell is built for the kind of workloads companies now want most: training and serving reasoning models that generate far more tokens and chain together multiple model calls. Nvidia said those workloads need more memory, faster chip-to-chip links, and larger rack-scale systems such as GB200 NVL72. (nvidia.com 1) (nvidia.com 2) Cloud availability has widened, but not enough to keep prices flat. CoreWeave said on April 15, 2025 that Cohere, International Business Machines, and Mistral were among the first customers on its GB200 systems, and Nvidia said Blackwell later became generally available in the cloud through CoreWeave. (investors.coreweave.com) (nvidia.com) Other public listings show how expensive the top tier has become. Lambda said eight-GPU B200 instances were available on demand in late 2025, while pricing trackers this spring showed single-GPU B200 rates around $5.98 an hour on RunPod and $6.08 on Lambda, with multi-GPU CoreWeave clusters priced higher in total. (lambda.ai) (deploybase.ai) The Ornn index itself has become more visible on Wall Street. Ornn said on April 2, 2026 that its Compute Price Index was added to the Bloomberg Terminal after six months as a live index tracking standardized GPU prices. (morningstar.com) Nvidia has been telling investors demand is still outrunning supply. At the company’s March 16, 2026 developer conference, Chief Executive Jensen Huang said he saw $1 trillion in orders for Blackwell and Vera Rubin systems through 2027. (cnbc.com) For companies building on rented chips instead of owned clusters, the latest signal is simple: the fastest artificial intelligence hardware is available, but each hour on it is getting more expensive. (techmeme.com) (the-decoder.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.