According to reports, the supply tightness of artificial intelligence (AI) chips is showing signs of easing, with companies that purchased large quantities of NVIDIA H100 80GB processors now attempting to resell them.
Delivery times for NVIDIA's H100 GPU, used for AI and high-performance computing (HPC) applications, have significantly shortened from 8-11 months to 3-4 months.
Some companies are reselling their H100 GPUs or reducing orders as the scarcity of these chips begins to diminish, and the costs of maintaining unused inventory are high.
This marks a significant shift from a year ago when acquiring NVIDIA's Hopper GPU was a major challenge.
The easing of AI processor supply shortages is also reflected in the increased ease of leasing NVIDIA's H100 GPU from cloud service providers such as AWS, Google Cloud, and Microsoft Azure.
For example, AWS has introduced a new service allowing customers to schedule shorter GPU leases, addressing previous chip availability issues and reducing the wait time for acquiring AI chips.
Despite improved chip availability and significantly reduced delivery times, demand for AI chips still far exceeds supply.
Companies that develop and train large-scale language models still face supply issues, largely due to the massive number of GPUs they require. These companies continue to experience delays of several months in obtaining the required processors or capacity.
As a result, prices for NVIDIA H100 and other processors have not decreased, and the company continues to enjoy high profit margins.
However, with the emergence of numerous alternatives to NVIDIA processors, such as those from AMD and AWS, the market may become more balanced.
Another reason is that companies have become more cautious in their spending on AI processors.
Nevertheless, as of now, the market demand for AI chips remains strong, and with large language models growing in size, the demand for computational performance is also increasing.
