Blackwell GPU prices spike

Hourly rental rates for Nvidia Blackwell GPUs have jumped sharply in recent months, rising to about $4.08 an hour — roughly 48% higher than two months ago — as demand for agentic AI workloads accelerates. Vendors and partners are also validating Blackwell across edge and industrial form factors, with firms announcing support for RTX Pro Blackwell in rugged edge systems and new Blackwell-based cloud offerings. (intellectia.ai, prnewswire.com)

Renting Nvidia’s newest Blackwell chips in the cloud now costs about $4.08 an hour, up from $2.75 two months earlier. (techmeme.com) That price jump comes from the Ornn Compute Price Index, which tracks graphics processing unit rental rates across cloud and on-premise markets and was added to the Bloomberg Terminal on April 2, 2026. (aijourn.com) Blackwell is Nvidia’s latest data-center chip platform, built for “inference,” the stage when trained artificial intelligence models answer questions and generate outputs for users. Nvidia says Blackwell is designed for “AI factories” running the largest models at scale. (nvidia.com, blogs.nvidia.com) The recent squeeze is tied to “agentic” artificial intelligence work, where software systems take multi-step actions instead of returning one reply. Nvidia said in February that Blackwell lowered token costs for open-source models, and in March it promoted new systems for local agentic artificial intelligence. (blogs.nvidia.com, blogs.nvidia.com) Cloud providers are still expanding supply. Nvidia said CoreWeave made Blackwell generally available in the cloud, and Boost Run said on April 13, 2026 that it achieved Nvidia’s “Exemplar Cloud” validation on Blackwell architecture systems. (blogs.nvidia.com, prnewswire.com) Blackwell is also moving beyond giant data centers. Supermicro said in March 2026 that it was expanding systems built around Nvidia RTX Pro 4500 Blackwell Server Edition graphics processing units for enterprise data centers and edge deployments with tight space, power, and cooling limits. (prnewswire.com) Nvidia had already laid groundwork for that push in August 2025, when it said Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, and Supermicro would ship enterprise servers using RTX Pro 6000 Blackwell Server Edition graphics processing units. (nvidianews.nvidia.com) The result is a market where new Blackwell capacity is arriving, but demand is rising fast enough that hourly rental prices are still climbing. For companies building artificial intelligence products, the chip bill is becoming a moving target instead of a fixed input cost. (techmeme.com, aijourn.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.