IT Home, October 24: Although NVIDIA has denied it, rumors of the RTX 5090 being discontinued have triggered price fluctuations in Huaqiangbei.
According to 21st Century Business Herald today, the channel prices for high-end models in the RTX 50 series are changing daily; the flagship RTX 5090, with an initial official retail price of approximately 15,000 to 16,000 yuan, has been inflated to 55,000 to 60,000 yuan, while the RTX 5080 has risen from an initial price of about 8,000 yuan to 17,000 yuan.
Xiong Rongrong, manager of Jinsuoli Computer, who has been in the Huaqiangbei industry for 16 years, said that two weeks ago, the channel price of the RTX 5080 was around 12,000 yuan, but it has risen sharply by 4,000 to 5,000 yuan in a short period; the RTX 5090, previously around 30,000 yuan, has surged directly to the 50,000-yuan level. Meanwhile, the mainstream model, the RTX 5060, has increased from approximately 2,300 yuan to over 3,600 yuan. Price differences for the same model within a single day or between consecutive days can reach hundreds of yuan, causing merchants to hesitate even to provide written quotes.
There was previously speculation that NVIDIA might halt supply of the GB202 chip for the RTX 50 series to redirect capacity toward RTX PRO professional GPUs, but this claim has been officially denied by NVIDIA.
IT Home previously reported that Jeff Fisher, Senior Vice President of NVIDIA's Personal Computing Business, responded after attending the RTX Spark launch event in Taipei on October 17, stating that NVIDIA has not discontinued the GeForce RTX 5090 series and will continue supplying it; however, he acknowledged that production capacity is indeed constrained, and supply shortages are a global issue facing the entire industry, with NVIDIA committed to maximizing product availability and continuing shipments.
A more accurate description than "discontinued" comes from Ada Keta, Vice President of Developer Relations at NVIDIA, who told media after the RTX Spark launch in Seoul that the root cause of the consumer GPU shortage is insufficient memory supply, "something NVIDIA cannot control."
According to EET-China, GDDR7 memory used in consumer graphics cards shares advanced manufacturing capacity with high-bandwidth memory (HBM) required for AI servers. Samsung, SK Hynix, and Micron—the three leading memory manufacturers—have allocated 70% to 90% of their advanced capacity to higher-margin HBM. By 2026, AI data centers are expected to account for approximately 70% of global memory production capacity.