Recently, NVIDIA has enforced its third across-the-board price hike for consumer-grade graphics cards within this year, in response to a significant spike in upstream memory costs, with the maximum increase hitting 30%. This price adjustment encompasses the full spectrum of GPU chips and GDDR memory modules, notably including mainstream mid-to-low-end models, thus drawing widespread market attention. Taking the RTX 5070 Ti as an illustration, its price has skyrocketed by 48.6% over the past six months, with the retail price surpassing 10,000 yuan. The price surge is directly attributable to Samsung DRAM memory price quote (Note: 'quotation' remains untranslated here as it seems to be a placeholder or error; if referring to the 'price quote', it is used correctly in this context) climbing by approximately 20%, resulting in a substantial rise in memory procurement costs—the price of 3GB particles has soared more than three times compared to the 2GB specification.
Consequently, manufacturers like ASUS have been compelled to cancel paid orders, demanding that users cover the price differential to finalize their purchases. For example, following the cancellation of RTX 5090 orders, the price jumped by nearly 12.5% upon its relisting. Panic-induced removals and hoarding by retail channels have further intensified supply shortages, underscoring the vulnerability of the global supply chain.
This price escalation not only imposes a financial strain on consumers but also necessitates AI computing power companies to reassess the expenses associated with cluster construction, prompting a shift towards more adaptable leasing models for computing power services.