TL;DR
Get business pricing on networking and server gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Data from Ornn Data, reported by a16z, shows rental rates for several NVIDIA GPUs rose from Jan. 1, 2026, with the H200 up 143% and the A100 up 9%. The figures track prices, not the reasons for the increases; demand, supply, data-center capacity and power constraints are not broken out.
Rental rates for several NVIDIA GPUs rose sharply in 2026, with the H200 increasing 143% from its Jan. 1 level, according to Ornn Data figures reported by a16z. The figures also show gains for the B200, RTX PRO 6000, H100 and A100, but do not identify how much of the movement came from AI demand versus limited supply or other constraints.
The reported increases vary by model. Since Jan. 1, 2026, the B200’s rental rate rose 111%, the RTX PRO 6000’s rose 73%, the H100’s rose 54%, and the A100’s rose 9%. The H200, a model launched in 2024, recorded the largest reported increase. These percentages describe rental-rate changes against the stated January starting point, not GPU sales prices or returns earned by owners.
The trend was not uniform throughout the year. The data summary says most tracked GPUs reached lower points in the first quarter. The H100 and B200 fell roughly 23% below their January levels before rebounding in April and May. The H100 remained broadly flat until early May, then rose over a period of weeks. The H200 climbed sharply in late May, surrendered about a quarter of that rise within weeks, and then resumed climbing into October.
The article connects the wider upward trend with a sharp rise in AI token consumption during spring, but the figures do not establish that as the cause of the rental changes. The data summary does not specify the source of token-use measurements or identify whether demand came from enterprise deployments, AI agents, or other workloads. It also gives no detailed breakdown of GPU availability, data-center capacity or power limits.
Higher Rental Rates Reshape GPU Economics
Rising rental rates can make deployed GPUs more valuable to their owners and raise the cost of computing for organizations that rent rather than buy hardware. The reported increases may matter to AI developers weighing cloud access against building or expanding their own infrastructure, though the data does not provide enough detail to calculate actual operating costs, profitability or customer impact.
The gains across different generations also point to a market in which older accelerators may still have a role. The report says the Tesla V100, an older NVIDIA GPU, is finding use for medium-sized language models. That suggests demand is not necessarily limited to the newest hardware, but the available information does not quantify V100 rental rates or show how widely it is being deployed.
For buyers and renters, the headline percentage alone is not a full pricing guide. Rental terms, memory capacity, location, availability and the duration of a contract could all affect the price a customer actually pays. None of those details is supplied alongside the summarized rate movements.
NVIDIA GPU server workstation for AI model inference
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Older GPUs Find New AI Workloads
The figures are presented against a backdrop of rising use of AI services, which consume computing resources when models process requests and generate tokens. The report associates higher token consumption in spring with the broad rise in rental rates, but does not provide a causal analysis. A correlation in the chart does not, by itself, show that token use drove the price changes.
The Tesla V100 is cited as an example of older hardware retaining utility. The article says that four V100 cards on an NVLink board can outperform the M5 Ultra and DGX Spark by as much as 230%, but the comparison’s test conditions and benchmark details are not included in the material summarized here. It also says higher-VRAM V100 variants are being sold at a $200 premium, without specifying the compared configuration, seller sample or measurement period.
Those points are distinct from the rental-rate series. The V100 example concerns reported performance and hardware premiums; the percentage changes for the A100, H100, H200, B200 and RTX PRO 6000 refer to rental rates measured from Jan. 1, 2026. The available account does not provide the underlying chart data or a full explanation of how the rates were collected.
NVIDIA H100 GPU server for machine learning
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Demand and Supply Drivers Remain Unspecified
The figures do not explain what drove the rental increases. Although the report links them to rising AI token consumption, it does not distinguish among enterprise adoption, agentic workloads, or other uses, and it does not establish a direct causal relationship.
Supply-side details are also missing. There is no breakdown of GPU inventory, data-center capacity, power availability, rental locations, contract lengths or pricing methodology. Those factors could affect rates, but the available information does not show whether or how much they contributed. The reported percentages therefore establish a pattern in the tracked rates, not a complete account of the market or a forecast of future prices.
Several comparisons also lack enough detail to independently evaluate them. The V100 performance claim does not include benchmark conditions, and the $200 premium for higher-VRAM models is not tied to a defined sample or timeframe. The number of listings or transactions behind that observation is not stated.
NVIDIA A100 GPU server for AI workloads
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
More Market Detail Is Needed
The next useful update would provide the underlying time series and explain how rental rates were sampled, including whether figures represent quoted prices, completed rentals or a particular set of providers. A model-by-model account of availability and contract terms would help readers assess whether the reported increases are broadly representative.
Further reporting would also need to separate demand indicators from supply constraints. Details on AI workloads, data-center capacity, power availability and hardware inventory could clarify why the rates moved and whether the increases persisted beyond the periods described. For now, the reported figures show substantial gains for several NVIDIA GPUs through October, while the causes and durability of those gains remain uncertain.
Source: Wccftech
GPU cloud computing server rental for AI training
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Which NVIDIA GPU had the largest reported rental-rate increase?
The H200 had the largest increase, rising 143% from its Jan. 1, 2026 level in the figures reported by a16z.
How much did the other listed GPU rental rates rise?
The reported increases were 111% for the B200, 73% for the RTX PRO 6000, 54% for the H100 and 9% for the A100, measured against Jan. 1, 2026.
Do the figures prove that AI token use caused rental prices to rise?
No. The report notes rising AI token consumption in spring alongside higher rates, but does not establish causation or identify which workloads drove demand.
Why is the Tesla V100 mentioned?
The report says the older V100 is finding renewed use for medium-sized language models. It also cites a performance comparison and a $200 premium for higher-VRAM variants, but does not provide the test conditions or pricing sample needed to assess those claims fully.
Source: Wccftech
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
