Hyperscalers List GPUs at Nearly 2× the GPU-Cloud Price, Fluence Research Finds
A new report reveals that hyperscalers list GPU prices at nearly double the median GPU-cloud price, with commitment lengths significantly affecting costs.
Fluence Research has published The AI Compute Race report, which reveals that hyperscalers list GPU prices at 1.98 times the GPU-cloud median price. Across H100, H200 and B200 models, the pricing varies from 1.4 times on the B200 to 2.8 times on the H100. For instance, the same H200 model lists from $4.08 to $10.60 per hour depending on the provider.
The report, produced with research partners including NEAR AI, io.net, Boundless, Phala, Gensyn, Fortytwo, Theta, Compute Labs and GAIB, demonstrates that where a company buys a GPU-hour significantly impacts the price. Commitment length matters almost as much as provider choice.
Against each provider's on-demand rate, a one-year reservation cuts the median price by 21%, while a three-year term reduces it by 48%. Spot capacity lists at just 53% of on-demand prices, offering savings for workloads that can be interrupted.
Rising GPU Costs
Buyers expecting GPU prices to fall may need to adjust their budgets. Seventeen of the 22 GPU models tracked by GetDeploying now cost more than a year earlier, with median H100 rents increasing by 13% to 30% depending on the measurement period.
The report identifies several factors driving these price increases. Demand has shifted toward open models, with Chinese models' share of OpenRouter tokens among companies in the US, Europe and the Global South rising from 6-13% to 57-67% in under eight months. This allows any company with GPUs to serve near-frontier models on controlled hardware.
The four largest hyperscalers are guiding $720-740 billion in 2026 capital spending, while the four largest disclosed deals by AI labs and Meta with neoclouds total approximately $233 billion. Much of this capacity is contracted years ahead, limiting availability.
Open-weight models are close behind the best closed ones, and companies are switching to them. Any company with GPUs can now serve a near-frontier model on hardware it controls. The GPUs are the hard part.
Financialization of GPUs
The GPU-hour is increasingly becoming a financial asset. Major exchanges like CME, ICE and Nodal Exchange have announced GPU futures contracts, each settling on different price indices. Meanwhile, lenders are already providing financing against GPU assets, with Lambda's investment-grade-rated loan priced at 6.78% and Sharon AI's, with no rating disclosed, at 9.95%.
The report concludes with four strategies for reducing compute costs: comparing the same chip across different provider types, committing when workloads are steady, matching GPU generation to model requirements, and looking beyond hyperscalers since much of the newest neocloud capacity is already contracted to labs years in advance.
The full AI Compute Race report is available at fluence.ai/compute-race-2026. Fluence Cloud, founded in 2017 by Dmitry Kurinskiy, Evgeny Ponomarev and Tom Trowbridge, has raised $14 million from investors including 1KX, Multicoin, Distributed Global, Blockchange, Tiger Global and Protocol Labs.