{"name":"Macrostack GPU Rental Price Index","description":"Published on-demand rental rates for AI accelerators, surveyed across neoclouds, marketplaces, serverless inference platforms and hyperscalers. The durable finding is the spread: the identical accelerator rents for many times more from one seller than another.","url":"https://www.macrostack.net/compute/gpu-prices","license":"https://creativecommons.org/licenses/by/4.0/","attribution":"Data from Macrostack (https://www.macrostack.net), CC-BY-4.0. Attribution with a link is a condition of the licence.","observed":"2026-09-09","unit":"USD per GPU-hour, on-demand unless the tier says otherwise","basis":"Published on-demand list rates, surveyed from third-party pricing indices and provider price pages. Not measured by Macrostack. Committed, reserved and negotiated rates are materially lower than every figure here and are not published by anyone.","disclaimer":"Macrostack does not rent clusters and does not benchmark them. Every figure here is a published list rate read from the cited source on the observation date. Committed, reserved and negotiated rates are materially lower than every figure in this file and are not published by anyone.","chips":[{"chip":"NVIDIA H100 SXM 80GB","slug":"h100-sxm-80gb","unit":"USD per GPU-hour, on-demand unless the tier says otherwise","lowUsdPerGpuHour":"1.38","highUsdPerGpuHour":"12.29","spreadMultiple":8.9,"spreadNote":"Roughly a 9x spread for a physically identical accelerator. This is the single most important number on this page: the chip is not the variable, the counterparty is.","specPage":"https://www.macrostack.net/silicon/nvidia-h100","observations":[{"band":"Marketplace and boutique floor","usdPerGpuHour":"1.38 - 1.49","tier":"on-demand, marketplace","note":"Brokered capacity. The machine you get is not a machine you chose, and interconnect quality is the thing that varies.","source":"https://intuitionlabs.ai/articles/h100-rental-prices-cloud-comparison"},{"band":"Specialist neocloud and marketplace on-demand","usdPerGpuHour":"2.00 - 3.00","tier":"on-demand","note":"Where RunPod and Vast.ai sit for a single H100 SXM. Spot listings dip below this.","source":"https://intuitionlabs.ai/articles/h100-rental-prices-cloud-comparison"},{"band":"Broad market median","usdPerGpuHour":"3.38","tier":"on-demand","note":"Roughly half what the same rate was a year earlier — the clearest evidence that H100 capacity stopped being scarce.","source":"https://shattered.io/h100-h200-b200-cloud-gpu-pricing-2026/"},{"band":"Hyperscaler on-demand ceiling","usdPerGpuHour":"11.68 - 12.29","tier":"on-demand, hyperscaler list","note":"Nobody signing a real contract pays this. It is a list price that exists to make the committed rate look like a discount.","source":"https://intuitionlabs.ai/articles/h100-rental-prices-cloud-comparison"}]},{"chip":"NVIDIA H200 SXM 141GB","slug":"h200-sxm-141gb","unit":"USD per GPU-hour, on-demand unless the tier says otherwise","lowUsdPerGpuHour":"3.59","highUsdPerGpuHour":"4.59","spreadMultiple":1.3,"spreadNote":"The tightest spread of any part here, because H200 supply arrived into a market that had already learned what H100 was worth.","specPage":"https://www.macrostack.net/silicon/nvidia-h200","observations":[{"band":"RunPod community cloud","usdPerGpuHour":"3.59","tier":"community","note":"Held at roughly this rate through most of 2026. Community capacity is other people's hardware.","source":"https://gpusmith.com/articles/en/gpu-cloud-rental-prices-h100-h200"},{"band":"RunPod secure cloud","usdPerGpuHour":"4.39 - 4.59","tier":"secure","note":"The same chip in a datacentre the provider controls. The delta is the price of knowing where your data physically is.","source":"https://gpusmith.com/articles/en/gpu-cloud-rental-prices-h100-h200"}]},{"chip":"NVIDIA B200","slug":"b200-180gb","unit":"USD per GPU-hour, on-demand unless the tier says otherwise","lowUsdPerGpuHour":"3.44","highUsdPerGpuHour":"16.11","spreadMultiple":4.7,"spreadNote":"A 4.7x spread, and the widest in absolute dollars. Blackwell capacity is still being allocated rather than sold, which is what a wide spread looks like from the outside.","specPage":"https://www.macrostack.net/silicon/nvidia-b200","observations":[{"band":"Vast.ai","usdPerGpuHour":"3.44","tier":"marketplace","note":"The floor of the surveyed market.","source":"https://intuitionlabs.ai/articles/data-center-gpu-pricing-2026"},{"band":"Lambda","usdPerGpuHour":"4.99 - 5.29","tier":"on-demand","note":null,"source":"https://intuitionlabs.ai/articles/data-center-gpu-pricing-2026"},{"band":"RunPod","usdPerGpuHour":"5.89","tier":"on-demand","note":null,"source":"https://intuitionlabs.ai/articles/data-center-gpu-pricing-2026"},{"band":"Median across surveyed providers","usdPerGpuHour":"6.11","tier":"on-demand","note":null,"source":"https://intuitionlabs.ai/articles/data-center-gpu-pricing-2026"},{"band":"Google Cloud","usdPerGpuHour":"16.11","tier":"on-demand, hyperscaler list","note":"The ceiling of the surveyed market, and 4.7x the floor for the same part.","source":"https://intuitionlabs.ai/articles/data-center-gpu-pricing-2026"}]}],"findings":[{"claim":"Neoclouds and marketplaces price 40% to 400% below the three major hyperscalers for the same accelerator.","source":"https://www.spheron.network/blog/gpu-cloud-pricing-comparison-2026/"},{"claim":"H100 on-demand rates roughly halved during 2026, while Blackwell-class parts held a wide spread — the signature of a market where the previous generation has become a commodity and the current one is still allocated.","source":"https://shattered.io/h100-h200-b200-cloud-gpu-pricing-2026/"}]}