The competition between AMD and NVIDIA in the artificial intelligence (AI) chip market has intensified. Despite AMD's recent launch of the new AI accelerator, MI325X, which is aimed at challenging NVIDIA's market dominance, the market response has been underwhelming. Following the announcement, AMD's stock price fell by approximately 4%, marking the largest intraday and closing decline since September.
On October 10, 2024, AMD unveiled the MI325X AI accelerator at an AI conference in San Francisco. This chip utilizes the same architecture as last year's MI300X; however, it is equipped with entirely new memory technology designed to enhance AI computing speed. Specifically, the MI325X boasts a memory capacity of 256GB, with bandwidth increasing from 5.3TB/s to 6TB/s, and a power consumption of up to 1000W, which represents an increase of 750W from the MI300X.
Although AMD CEO Lisa Su stated that the performance of the MI325X surpasses that of NVIDIA's H100, the market response has been disappointing. Analysts point out that investor expectations for AMD primarily focus on whether its AI business can yield substantial financial returns, rather than merely on the release of new products.
(Source: uSMART HK)
NVIDIA continues to maintain a strong leadership position in the AI chip market. According to the latest data, NVIDIA's data center revenue reached $22.6 billion in the first quarter of fiscal year 2025, setting a record for the highest quarterly income from this segment, with a quarter-over-quarter increase of 23% and a year-over-year surge of 427%. In contrast, AMD's data center business reported net revenue of only $2.8 billion during the same quarter, despite achieving a year-over-year growth of 115%.
Moreover, NVIDIA's H100 chip has demonstrated strong market performance, becoming the preferred AI processor for many large technology companies, including Microsoft and Meta. Analysts project that NVIDIA's data center revenue will reach $110.36 billion in 2024, while AMD's anticipated revenue is only $12.83 billion.
|
Metric |
AMD MI325X |
NVIDIA H100 |
|
Memory Capacity |
256GB |
80GB |
|
Memory Bandwidth |
6TB/s |
3TB/s |
|
Power Consumption |
1000W |
700W |
|
Processing Power (FP16) |
64 TFLOPS |
60 TFLOPS |
|
Process Node |
5nm |
4nm |
|
Projected Market Share Growth |
15% |
35% |
(Source: uSMART HK)
From the table, it is evident that AMD has advantages in memory capacity and bandwidth; however, NVIDIA excels in power consumption and process technology. Although the MI325X slightly surpasses the H100 in processing capability, the latter's market acceptance and practical application performance remain dominant.
Despite AMD's increasing investment in AI chips, market analysts generally believe that AMD will require time to catch up with NVIDIA. While the MI325X represents a technological breakthrough, its ability to surpass NVIDIA's products in real-world applications remains uncertain. Additionally, AMD's management must demonstrate a clear impetus for market share growth in order to regain investor confidence.
Furthermore, Lisa Su indicated at the conference that by 2028, the market for AI data center GPUs is expected to reach $500 billion, growing at an annual rate of over 60%. This market potential has attracted considerable interest from various enterprises, and AMD hopes to secure a share in this burgeoning sector.
The competition between AMD and NVIDIA in the AI chip arena continues. Although AMD is consistently innovating technologically, market reactions and financial performance indicate a significant gap between the two companies. Whether AMD can narrow this gap in the future will depend on its products' market acceptance and its ability to achieve financial breakthroughs.
As AI technology continues to evolve, competition in the AI chip market is expected to intensify, with investors and industry observers closely monitoring the developments of both companies. Despite facing challenges, AMD's efforts and innovations may yield unexpected breakthroughs in the future market landscape.
After logging into the uSMART HK app, click on the "Search" option located at the top right of the page. Enter the stock code, such as " AMD " to access the details page where you can review transaction details and historical trends. Then, click on the "Trade" option at the bottom right corner, select the "Buy/Sell" function, and finally fill in your transaction conditions before submitting your order.
(Source: uSMART HK)
