Dear Users,
As China's GPU cloud compute market continues to expand rapidly, spot prices for GPU cloud compute resources are experiencing frequent fluctuations. Market participants increasingly demand a standardized and transparent GPU cloud compute spot price benchmark to better track compute cost trends and reduce information asymmetry.
To fill the market gap for a domestic GPU cloud compute spot price index, SMM has completed extensive market research, methodological validation, and technical development. Effective July 3, 2026, SMM will officially launch two new data series — the SMM GPU Cloud Compute Spot Price Index and the SMM GPU Cloud Compute Spot Listing Price — updated on a daily basis on every business day.
1. Overview of New Data Points
The new data points are organized into two series, covering multi-dimensional price references across China's public GPU cloud compute spot market:
2. SMM GPU Cloud Compute Spot Price Index (Series A)
2.1 Product Positioning
The SMM GPU Cloud Compute Spot Price Index is a standardized spot price index derived from actual market transaction data and representative public listings across mainstream GPU cloud compute platforms in China. Through specification normalization, weighted cleaning, and quantile calculation, SMM produces a reliable reference price that reflects the real-time market-clearing level of China's GPU cloud compute spot market.
2.2 What Is "GPU Cloud Compute" (算力租赁)?
In China's market, "算力租赁" (GPU Cloud Compute) is the established industry term for commercial GPU compute services. In this announcement, it refers to the full spectrum of commercial GPU cloud compute offerings available in the Chinese market — where buyers obtain time-limited access to GPU hardware on a recurring fee basis, regardless of contract length.
The term "算力" (compute power) is China-specific and carries the connotation of "temporary use of computational resources in exchange for payment." Internationally, this market is commonly referred to as:
-
GPU Cloud Compute — the most widely used international term
-
GPU-as-a-Service (GPUaaS) — industry research and analyst standard
-
Cloud GPU — neocloud provider preference (CoreWeave, Lambda, RunPod, etc.)
These terms are equivalent to "算力租赁" as used in this announcement.
2.3 Coverage
2.4 Index Naming Convention
The index is published under a standardized naming convention aligned with international GPU cloud compute market practice:
SMM-{GPU Model}-{VRAM}-{Instance Type}-{Region}-{Billing Tier}
Example: SMM-4090-24G-Multi-GPU Node-YRD-On Demand
Model Summary Level: SMM-4090-On Demand-ALL (aggregated across instance types and VRAM variants; for reference only)
3. SMM GPU Cloud Compute Spot Listing Price (Series B)
3.1 Product Positioning
The SMM GPU Cloud Compute Spot Listing Price is a companion data series to Series A, providing direct market listing price references. It presents representative platform real-time public listing prices under each data point, enabling market participants to directly compare current posted rates across providers.
3.2 Coverage
3.3 Price Specifications
4. Usage Guide and Important Notes
4.1 Recommended Use Cases
4.2 Important Notes
-
These data points represent public GPU cloud compute spot market listing prices, and do not cover defense/military sector, specialized supercomputing, or internal custom compute pricing.
-
The price range (B-Low to B-High) reflects the distribution of publicly posted listing prices across platforms. Actual transaction prices may fall outside this range depending on real-time supply & demand dynamics, buyer negotiation leverage, and specific contract terms.
-
Instance Type Summary and Model Summary aggregated data offer higher coverage rates and are recommended as the primary reference. Full price point level data may have higher missing data rates; refer to the methodology review document for details.
-
SMM will continuously optimize source coverage and calculation methodology, with gradual expansion of additional GPU models and regions.
-
For data corrections, price verification inquiries, or new data source recommendations, please contact the SMM Compute Research Team.
5. Reference: International GPU Cloud Compute Pricing Terminology
This section provides a glossary of key English pricing terminology used in the global GPU cloud compute market, aligned with AWS, Google Cloud, Azure, and major neocloud standards. Note that "算力租赁" (GPU Cloud Compute) is China-specific industry terminology; the equivalent international terms are listed below:
6. Contact Information
For inquiries, data feedback, or methodology questions, please contact the SMM Compute Research Team:
-
Email: fangwenda@smm.cn; liuxiaolei@smm.cn
-
Phone: +86 021-51666821
Shanghai Metals Market (SMM) Compute Research Team June 25, 2026