Swift Delivery: NVIDIA H100 GPU Cycle Now 3-4 Months

In recent international reports, it has been highlighted that NVIDIA is experiencing a noticeable reduction in the delivery cycle for its leading artificial intelligence (AI) chip, the H100, due to improvements in market supply conditions. Previously, the waiting period for users ranged from 8 to 11 months, but it has now shortened substantially to 3 to 4 months. Notably, major cloud service providers such as AWS, Google Cloud, and Microsoft Azure are contributing to this shift by offering convenient AI computing services. Consequently, enterprises that were once prominent purchasers of H100 GPUs are now entering the resale market for these products.

The report underscores a decline in the scarcity of H100 GPUs, coupled with the high maintenance costs, prompting certain clients to divest this hardware and migrate towards cloud-based solutions. This represents a significant departure from the market's conditions a year ago.

Despite the increased accessibility of H100 GPUs, the demand for AI solutions continues to outpace supply, preventing a substantial drop in the price of H100. Notably, alternatives like the AMD MI300 have entered the market, offering superior cost-effectiveness.

According to the latest projections from research firm TrendForce, the global demand for high-end AI servers, powered by NVIDIA, AMD, or other top-tier ASIC chips, is anticipated to reach 150 million units in 2024. Key industry players, including Microsoft, Google, AWS, and Meta, are expected to collectively capture over 60% of the global AI server demand, with NVIDIA chips maintaining a dominant share of over 70%.

However, NVIDIA's leadership in the AI sector faces several challenges. Firstly, due to the U.S. technology export ban, the company's sales in the Chinese market are anticipated to be impacted. Additionally, the growing trend among major players such as Amazon AWS, Google, Microsoft, and Meta to develop proprietary ASIC chips is gaining momentum for reasons related to scale and cost. Furthermore, AMD's competitive edge, driven by its high cost-effectiveness, with products priced at 60% to 70% of NVIDIA's equivalents, is facilitating market penetration, particularly among flagship customers. Predictions suggest that Microsoft will emerge as the primary customer for AMD MI300 GPU solutions in 2024.