In a frenzy of demand for AI and HPC GPUs in China, Nvidia’s AI GPUs are fetching astronomical prices, with some units selling for as high as $70,000. This unprecedented surge in prices has been primarily fueled by the soaring demand for generative artificial intelligence-based services and growing apprehensions about potential restrictions on GPU sales for AI workloads by the U.S. government. As a result, Nvidia’s A800 and H800 compute GPUs, specifically tailored for China’s market with slightly reduced performance and scalability compared to their full-fledged counterparts, have become increasingly elusive, leaving potential buyers with limited options.
Rumors spark a 20% price surge
Last month, Nvidia’s A800 GPU, specially designed for AI and high-performance computing applications, saw its price surge by 20% overnight, reaching ¥110,000 ($15,000). These rapid price increases came in response to rumors suggesting that the U.S. government might impose export restrictions on AI and HPC products destined for China. Subsequently, the prices of Nvidia’s A800 and H800 compute GPUs have experienced further escalation, with units now fetching as much as ¥500,000 ($69,950). The GPU (Graphics Processing Unit) is a specialized processor made up of many smaller and more specialized cores. By working together, the cores deliver massive performance when a processing task can be divided up and processed across many cores. With NVIDIA AI software, including RAPIDS™ open-source software libraries, GPUs substantially reduce infrastructure costs and provide superior performance for end-to-end data science workflows.
Difficulties in procurement
Even with the exorbitant prices, obtaining Nvidia’s China-oriented A800 and H800 GPUs has proven to be a challenging task. Instead of purchasing from distributors or resellers, potential buyers are forced to reach out directly to Nvidia China or even the company’s corporate headquarters. The limited availability of these GPUs raises concerns for owners of AI services and clusters, as the inability to meet the surging demand could result in degraded service quality and potential loss of business over time. The US Commerce Department could move as soon as early next month to stop the shipments of chips made by Nvidia and other chip makers to customers in China and other countries of concern without first obtaining a license, the people said in late June.
Nvidia’s dominance in the AI cluster market
The overwhelming majority of AI clusters currently in use are built around Nvidia’s compute GPUs, leveraging the proprietary CUDA software layer exclusively compatible with the company’s processors. This reliance on Nvidia’s technology has further contributed to the massive demand for their GPUs, as customers seek to bolster their AI capabilities with cutting-edge hardware. However, the scarcity of these GPUs threatens to disrupt the smooth functioning of AI services, potentially hindering their further development.
Nvidia’s silence on AI GPU pricing
Nvidia has chosen to remain tight-lipped on the subject of pricing for its data center GPUs, leaving many to speculate about the validity of the reported prices. While the demand for GPUs in China remains undeniable, the company’s refusal to comment on the matter leaves some room for skepticism. Nevertheless, the steep price tag of an Nvidia H800 compute GPU in the U.S. stands at $30,603, implying that the Chinese market is experiencing even more intense price fluctuations.
CDW, one of the prominent resellers of data center hardware, lists only one of Nvidia’s compute GPUs, and the estimated delivery time spans between 5 to 7 business days. Such indications point to possible supply constraints, leading to scarcity and delayed deliveries. The constrained availability of these GPUs exacerbates the pricing situation, pushing prices even higher in response to limited stock.
As the demand for AI and HPC GPUs in China continues to soar, prices for Nvidia’s A800 and H800 compute GPUs have reached unprecedented heights, with units fetching up to $70,000. Fueled by fears of potential export restrictions from the U.S. government and the overwhelming reliance on Nvidia’s technology in the AI cluster market, these GPUs have become increasingly difficult to obtain. The future trajectory of GPU prices remains uncertain, but the implications of the supply shortage could pose significant challenges for businesses reliant on AI services and clusters in China.
Source: https://www.cryptopolitan.com/ai-gpu-china-price-hits-70000/