Nvidia to Gain due to Demand for AI Servers Despite Price Challenge from AMD

Microsoft, Google, AWS, and Meta are poised to collectively dominate over 60 percent of global demand for high-end AI servers, TrendForce said.
AI Server demand in 2024

Nvidia’s GPU-based servers are leading the charge.

The report underscores Nvidia’s ongoing challenges, particularly in development, as it grapples with US restrictions. Despite holding a dominant position in the data center sector, capturing up to 70 percent of the AI market with its GPU servers, Nvidia faces hurdles that could limit its growth trajectory.

Foremost among these challenges is the US ban on technological exports, which has prompted China to pursue self-reliance in AI chips, with Huawei emerging as a significant competitor. Nvidia’s China-specific solutions, such as the H20 series, may not match the cost-effectiveness of its flagship models, potentially eroding its market dominance.

Additionally, the trend toward proprietary ASIC development among major US cloud providers like Google, AWS, Microsoft, and Meta poses a significant threat. This trend, driven by scale and cost considerations, could further challenge Nvidia’s position. Moreover, AMD’s competitive pricing strategy, offering products at 60–70 percent of the prices of comparable Nvidia models, presents another obstacle to Nvidia’s dominance.

Nvidia is diversifying its product portfolio. The company plans to transition from the A100 series to the more advanced and costly H100 series in 2024, complemented by the introduction of the H200 series with superior HBM3e specifications.

Nvidia aims to balance cost and performance by aggressively pricing the H200 to appeal to cloud service provider clients. Additionally, Nvidia is expanding its market reach by pursuing partnerships with major players like Meta, Google, AWS, and OpenAI, targeting expansion into telecommunications, automotive, and gaming industries.

Looking ahead, Nvidia plans to unveil its next-gen B100 products in late 2024, expected to surpass the H series in efficiency. These products will feature a significantly higher HBM memory capacity, catering to the demands of high-performance computing or accelerating LLM AI training.

Nvidia also plans to introduce the L40S, designed for enterprise customers, and the L4, targeting cloud or edge AI inference applications, thereby broadening its offerings for mid-range and entry-level segments.

To address the GPU supply shortage experienced in 2023, Nvidia is increasing CoWoS and HBM production capacities. This proactive approach aims to cut the average delivery time in half by the second quarter of 2024, thereby alleviating supply chain bottlenecks that have hindered AI server availability.

Related News

Latest News

Latest News