Blackwell GPU rents spike
Hourly rental prices for Nvidia’s Blackwell GPUs have jumped sharply in recent weeks, with an index citing a rise to $4.08 per hour — up about 48% from two months ago. The move is linked to increased demand for agentic AI workloads, and vendors are positioning themselves around premium Blackwell access while Nvidia says it's using AI to speed chip design workflows. (intellectia.ai) (prnewswire.com) (tomshardware.com).
Renting Nvidia’s newest Blackwell graphics processors now costs about $4.08 an hour, up 48% from roughly $2.75 two months ago. (techmeme.com) The jump came from the Ornn Compute Price Index, which tracks market pricing for rented graphics processing units, or GPUs, used to train and run artificial intelligence systems. Techmeme’s roundup of the Wall Street Journal report said the increase was tied to rising demand for “agentic” artificial intelligence tools that carry out multistep tasks on their own. (techmeme.com) A GPU rental market works like short-term power for artificial intelligence: companies that cannot buy enough chips outright pay cloud providers by the hour instead. Blackwell is Nvidia’s latest high-end architecture, and current comparison sites show B200 pricing varies widely by provider, with some listed rates still well above the $4.08 index level. (getdeploying.com) (datastorage.com) Cloud vendors are using Blackwell access as a selling point while supply stays tight. Boost Run said on April 13 that it had earned Nvidia’s “Exemplar Cloud” validation on Blackwell, a designation it said places it alongside providers including CoreWeave, Nebius, Oracle Cloud Infrastructure, and Microsoft Azure. (prnewswire.com) (tmcnet.com) That vendor scramble follows a broader shift in artificial intelligence demand from chatbots toward software that keeps working after a prompt, calling tools, writing code, and checking results. The Decoder, citing the Wall Street Journal report, said that shift has coincided with outages, rationing, and higher compute prices across the sector. (the-decoder.com) Nvidia is also telling customers that artificial intelligence is changing the way chips are built inside the company. Tom’s Hardware reported Monday that Nvidia Chief Scientist Bill Dally said an internal model cut one graphics processor design task from 10 months of work by eight engineers to an overnight run, while adding that fully autonomous chip design is still far off. (tomshardware.com) Ornn said on April 2 that its compute price index had been added to the Bloomberg Terminal after six months as a live index, a sign that graphics processor pricing is starting to be treated more like a financial market benchmark. The company said the index was built to give institutional users standardized visibility into GPU prices. (aijourn.com) For now, the Blackwell market is sending a simple signal: the newest Nvidia capacity is scarce, and companies that need it immediately are paying more for every hour they can get. (techmeme.com)
