GPU Rental Market Update: September 17, 2026
The cheapest GPU with at least 24 GB VRAM is the Tesla V100 at $0.1356 per hour on vast.ai, with 13 offers available.
Current Cheapest Options by VRAM Tier
For users needing at least 24 GB VRAM, the Tesla V100 with 32 GB VRAM is the lowest-cost option at $0.1356 per hour from vast.ai, offering 13 listings. This same GPU also serves the 32 GB VRAM tier at the same price and offer count.
The next tier, requiring at least 48 GB VRAM, is served by the RTX A6000 with 48 GB VRAM at $0.33 per hour on runpod, though the number of offers is not specified in the data.
For 80 GB VRAM or more, the A800 PCIE with 80 GB VRAM is available at $0.4681 per hour from vast.ai with 3 offers. The RTX PRO 6000 Max-Q (96 GB VRAM) costs $1.2015 per hour on vast.ai with 21 offers, and the MI300X (192 GB VRAM) is priced at $2.39 per hour on runpod, with offer count unspecified.
Notable Price Changes vs. Yesterday
The RTX A4500 saw the largest increase, rising 181.1% from $0.0676 to $0.19 per hour on runpod. The RTX 4080S followed with a 162.4% jump from $0.1884 to $0.4944 per hour on vast.ai. The H200 increased 149.9% from $1.9751 to $4.9351 per hour on vast.ai.
The Tesla T4 rose 120.5% from $0.0674 to $0.1486 per hour on vast.ai. In contrast, the RTX 2060 dropped 71.4% from $0.2422 to $0.0692 per hour on vast.ai. The RTX PRO 4500 declined 40.2% from $0.72 to $0.4304 per hour on vast.ai, matching the week-ago change for this GPU.
Other notable decreases include the RTX 4070S Ti (-29.8%), A100 PCIE (-29.7%), and RTX 3060 Ti (-26.6%). The RTX 2070 increased 27.5% from $0.0684 to $0.0872 per hour on vast.ai. A total of 55 chips showed no change or changed within 1% compared to yesterday.
Weekly Trends and Market Shifts
Over the past week, the RTX 4080S increased 146.8% from $0.2003 to $0.4944 per hour on vast.ai, and the CMP 170HX rose 131.5% from $0.2134 to $0.4941 per hour on the same provider. The GTX 1660 Ti experienced the largest weekly drop, falling 77.5% from $0.3463 to $0.0778 per hour on vast.ai.
The RTX 2070 gained 58.5% over the week, rising from $0.055 to $0.0872 per hour on vast.ai. The A100 PCIE decreased 41.9% from $0.8082 to $0.4694 per hour on vast.ai, and the RTX 3090 Ti fell 41.1% from $0.27 to $0.1589 per hour on the same provider. The RTX PRO 4500 again showed a -40.2% change, consistent with its daily shift.
Six new chip types entered the listings since last week: GTX 1050 Ti, Quadro P2000, RTX 2080, RTX 4000 Ada SFF, RTX 4500Ada, RTX 5000Ada, and RTX 6000D. Seven were removed: Q RTX 4000, RTX 2060S, RTX 3070 laptop, Tesla P4, Tesla P40, Q RTX 5000, and Q RTX 6000. Forty chips remained unchanged or varied by less than 1% over the week.
Most Frequently Offered GPUs
The RTX 4090 appears twice in the most-offered list with 35 offers each on vast.ai, though at differing prices: $0.3072 and $0.4956 per hour. This duplication may reflect different configurations or time-stamped entries.
The H200 is listed twice with 26 offers each on vast.ai, priced at $1.9751 and $4.9351 per hour — likely representing different instance types or availability windows. The Titan Xp has 22 offers at $0.0406 per hour on vast.ai, making it the lowest-priced GPU among the most frequently listed options.
Key takeaways
- The Tesla V100 remains the most cost-effective GPU for workloads requiring 24 GB VRAM or more, at $0.1356 per hour.
- The RTX 4080S and H200 showed the largest hourly and weekly price increases, more than doubling in some cases.
- The GTX 1660 Ti suffered the steepest weekly decline, dropping over three-quarters in price.
- The RTX 4090 and H200 each appear multiple times in the high-offering list at different price points, suggesting variant offerings.
- Despite volatility in individual GPUs, 55 chips were stable versus yesterday and 40 versus last week, indicating broader market steadiness.