🌐 ▾Simplified ChineseTraditional ChineseENJapanese한국어

GPU Compute Price Monthly Report for 2026/10

GPU Compute Radar · Monthly release · Issue 2026-10 · Data updated: 2026-11-12
⚡ Straight answer: Based on daily snapshots of publicly listed prices across 31 compute platforms (including 13 based in mainland China), this site ingested 2026 entries during 10/166;\n Among 7 comparable GPU models, median pay-as-you-go prices rose for 4 models, fell for 0, and stayed flat (±3%) for 3.\n % change is calculated as ‘baseline = median of first 3 days of month’ vs. ‘full-month median’, counting only pay-as-you-go listings — spot and monthly plans excluded. Data licensed under CC BY 4.0 — free to cite.

I. GPU Model Price Change Ranking for 2026/10 (pay-as-you-go median hourly rate, MoM baseline)

Baseline covers 2026-09-28–2026-09-30, a total of 3 snapshot days (daily snapshots available only from 2026-09-28 onward; the first baseline period is shorter and fully disclosed); report period covers 2026-10-01–2026-10-31, a total of 31 snapshot days. A GPU model must have at least 4 quoted prices to be included in comparison.

Top Gainers

GPU modelBaseline medianMedian for 2026/10Month-on-Month ChangeNumber of platforms selling
A100 80GB¥8.39¥10.74+28.0%15
A6000 48GB¥2.53¥2.85+12.9%7
RTX 5090¥2.68¥2.99+11.6%11
A800 80GB¥6.40¥6.80+6.2%5

Top decliners

No GPU model dropped more than 3% this period (all comparable models meeting the sample threshold remained flat or rose).

II. GHPI Price Index: Trend for 2026 Year 10 Month

MethodologyStart of monthEnd of monthChange
Overall domestic median¥5.15/hour¥5.15/hour+0.0%

(2026-10-01 → 2026-10-31, pay-as-you-go median price; 31 snapshot days within the month) Public list prices exhibit stickiness — platforms rarely adjust daily; prolonged index stability reflects genuine ‘no price change’, not unchanged transaction prices. Index methodology and daily fixing details at GHPI Price Index.

III. Domestic vs. international price gap (current observation)

The table below shows the current observed values at the time of publication: median hourly rental price for the same GPU model on domestic platforms vs. overseas platforms (only fresh pay-as-you-go quotes within the last 48 hours, with at least 2 platforms on each side). Price gaps are dynamic; real-time figures are available at Cross-Border Price Gap Dashboard.

GPU modelDomestic medianOverseas medianPrice spread
L40S 48GB¥5.03¥10.36Domestic prices are 51% cheaper
RTX PRO 6000 96GB¥10.90¥14.54Domestic prices are 25% cheaper
A100 80GB¥11.07¥10.61International prices are 4% cheaper

IV. Dedicated chapter: Domestic AI accelerators

Domestic GPUs—including Ascend, Hygon, and Kunlunxin—are excluded from overseas price-comparison sites and represent rentgpu.link’s deepest time-series asset. Current availability:

GPU modelMedian hourly rental priceCurrent Lowest PriceNumber of platforms selling
Ascend 910B2 64GB¥3.58/hourExtreme Intelligence Compute ¥2.972
Ascend 910C 128GB¥13.20/hourShared compute ¥13.201
Hygon BW1000 64GB¥7.00/hourShared compute ¥7.001
Kunlunxin P800 96GB¥8.50/hourShared compute ¥8.501

V. Methodology and How to Use This Report

Citations and data download

This report’s data is licensed under CC BY 4.0 — attribution required for reuse:
GPU Compute Price Monthly Report for 2026/10 | GPU Compute Radar rentgpu.link (updated 2026-11-12) | https://rentgpu.link/posts/2026-10-gpu-price-monthly | CC BY 4.0

Monthly frozen data (permanent citation link): /data/v/2026-10.json · Daily snapshots (open dataset page): /data/snapshots/ · Price index: /data/ghpi.json · DOI (quarterly archive): 10.5281/zenodo.23868319

Rankings and conclusions are auto-generated from data; prices reflect those on each platform’s order page. This site’s rankings are unaffected by commercial partnerships; all advertising collaborations are clearly labeled “Promotion”.
🐆 Xiao Bao · AI AssistantPowered by Tongyi Qwen
Hello there, I’m Leopard~ Feel free to ask about GPU pricing, cost-effective rental strategies, or what Pro membership offers.