[SMM Hashrate Midday Review] 4090 lowest price in Pearl River Delta has continued to drop, new resource release lowers the lower bound.

Published: Aug 24, 2026 11:43
The lowest monthly rent for 4090 in the Pearl River Delta dropped to 7,200 yuan (down 1.37%), and the lowest card hour dropped to 1.25 yuan (down 1.57%). The reason was the release of new low-cost resources and the invalidation of previous quotes, while other prices remained unchanged. The 4090 is primarily used for local enterprise inference, and due to diversion by workstation-level substitutes and the continuous influx of new resources, the price center has shifted downward. Low-price downward probing may still continue, benefiting computing power buyers.

[Market Overview] Today, the lowest price in the 4090 Pearl River Delta region continued to edge down. For SMM-4090-24G-Complete Unit-Pearl River Delta-Monthly Package, the lowest monthly rent fell from 7,300 yuan to 7,200 yuan, and the lowest hourly rate fell from 1.27 yuan to 1.25 yuan. Prices for other models and regions remained unchanged, and the overall market was stable.

[4090 Pearl River Delta: New Supply Released, Lowest Price Continued to Decline] For SMM-4090-24G-Complete Unit-Pearl River Delta-Monthly Package, the lowest monthly rent fell from 7,300 yuan to 7,200 yuan, down 100 yuan, a monthly rent decline of 1.37%; the lowest hourly rate fell from 1.27 yuan to 1.25 yuan, down 0.02 yuan, an hourly rate decline of 1.57%. This round of decline was driven by the release of new low-priced supply; the previous lowest listing prices of 7,300 yuan was consequently invalidated, while other prices remained unchanged. As the flagship consumer-grade model being repurposed for AI inference, the continued decline in the lowest price indicates that both incremental and existing supply in the region continued to flow in, and low-priced supply was taken up as soon as it entered the market.

[SMM Viewpoint] The continued decline in the lowest price for 4090 in the Pearl River Delta corroborated the view that the price center for consumer-grade cards repurposed for AI continued to shift downward. SMM believed two points warranted attention: first, the core demand for 4090 was enterprise on-premises deployment for large-model inference; this scenario was highly sensitive to cost performance, and the release of new low-priced supply could be transacted quickly, pushing the lowest-price anchor down step by step. Second, desktop-class AI workstations such as NVIDIA DGX Spark continued to divert demand; coupled with the continuous entry of new supply, 4090 rental prices lacked upward momentum, and the downward probing of low prices might still continue. For buyers, the bargain-hunting window for 4090 in the Pearl River Delta was expanding.

Data Source Statement: Except for publicly available information, all other data are processed by SMM based on publicly available information, market communication, and relying on SMM's internal database model. They are for reference only and do not constitute decision-making recommendations.

For any inquiries or for more information, please contact: lemonzhao@smm.cn
For more information on how to access our research reports, please contact:service.en@smm.cn
Related News
[SMM Computing Power News] East China 115 Units of 4090 Idle for Rent, Monthly Rent 7,100 First-Hand Resources
2 hours ago
[SMM Computing Power News] East China 115 Units of 4090 Idle for Rent, Monthly Rent 7,100 First-Hand Resources
Read More
[SMM Computing Power News] East China 115 Units of 4090 Idle for Rent, Monthly Rent 7,100 First-Hand Resources
[SMM Computing Power News] East China 115 Units of 4090 Idle for Rent, Monthly Rent 7,100 First-Hand Resources
SMM August 24 news: About 115 idle RTX 4090 servers in east China are available for lease, directly from the resource owner, with a monthly rent of 7,100 yuan per unit. Options include a lease term of 6-12 months, a minimum of 32 units, and a deposit of one month's rent plus one month's payment. Each machine features 8×4090 24G, dual Xeon 8352V, and 512GB DDR4. SMM believes that the rent of 7,100 yuan per unit with flexible splitting is competitive in both price and minimum lease threshold.
2 hours ago
[SMM Computing Power Express] Computing Power Demand Side Multi-Point Boosting: 950 Nodes Going Online in Batches
2 hours ago
[SMM Computing Power Express] Computing Power Demand Side Multi-Point Boosting: 950 Nodes Going Online in Batches
Read More
[SMM Computing Power Express] Computing Power Demand Side Multi-Point Boosting: 950 Nodes Going Online in Batches
[SMM Computing Power Express] Computing Power Demand Side Multi-Point Boosting: 950 Nodes Going Online in Batches
SMM learned that signals from the computing power demand side are intense: a computing-electricity collaboration project of a power generation enterprise and the expansion of a smart computing center of an operator are progressing steadily, while 950 nodes are being launched in batches (Q3 for cloud service providers, Q4 for operators), with supporting node demands from leading internet enterprises released simultaneously. SMM believes that the diversified growth on the demand side indicates that computing power construction has entered a stage of multi-subject collaborative volume release, which will support medium and long-term computing power leasing demand.
2 hours ago
[SMM Computing Power Express] Computing-Electricity Collaboration Model Works, Green Electricity Direct Supply Provides 760 Million kWh Annually
2 hours ago
[SMM Computing Power Express] Computing-Electricity Collaboration Model Works, Green Electricity Direct Supply Provides 760 Million kWh Annually
Read More
[SMM Computing Power Express] Computing-Electricity Collaboration Model Works, Green Electricity Direct Supply Provides 760 Million kWh Annually
[SMM Computing Power Express] Computing-Electricity Collaboration Model Works, Green Electricity Direct Supply Provides 760 Million kWh Annually
SMM learned that a power generation enterprise in North China has implemented a business model of coordinating green electricity projects with computing power. In the first phase, it supplies about 761 million kWh of green electricity annually, with the effective utilization rate of new energy power generation stable at over 95%. It adopts a model of direct supply of green electricity plus energy storage configuration (18% in the first phase, 25.4% in the second and third phases). SMM believes that the cost anchor for coordinating computing and electricity is taking shape, and direct supply of green electricity is expected to stabilize the electricity cost for computing power and strengthen the electricity price advantage of hub regions.
2 hours ago