AMD's AI Chip Sales Might Face Challenges in 2025

Advanced Micro Devices (AMD) has established itself as the second major player in the artificial intelligence (AI) accelerator market. However, it still trails significantly behind Nvidia. Currently, Nvidia reports approximately $30 billion in quarterly revenue from its data center operations, while AMD anticipates around $5 billion in AI accelerator sales for the entire year of 2024.

Despite the strong demand for AI accelerators, AMD's relatively small market share suggests that there is ample opportunity for the company to expand its AI business. Nonetheless, forecasts for 2025 indicate that it may not be a particularly strong year for AMD, based on insights from a Wall Street analyst.

Anticipated Slower Growth

An analyst from Wolfe Research recently downgraded AMD's stock to a “hold” rating and eliminated the price target, citing concerns over the growth of data center graphics processing unit (GPU) sales. The analyst highlighted the production plans of original design manufacturers as a basis for this increased skepticism.

The Wolfe analyst projects AMD’s data center GPU revenue will fall between $1.5 billion and $2 billion for the last quarter of 2024. For the year 2025, the forecast has been adjusted down to just $7 billion, which falls short of the $10 billion that many analysts expect. This figure indicates that AMD's revenue from data center GPUs may not see significant growth throughout the quarters of this year.

It is important to view analyst opinions with caution, as we are still early in the year and the AI sector is known for rapid developments. AMD has made advancements in AI accelerators and has begun attracting clients. Notably, Oracle is utilizing AMD’s MI300X GPUs within its cloud data centers, while smaller cloud provider Vultr has also chosen AMD GPUs to support its AI services.

An ongoing challenge for AMD, along with its competitors outside of Nvidia, is the established software ecosystem that Nvidia offers. Nvidia GPUs benefit from a proprietary software framework known as CUDA, which has been utilized extensively in both industry settings and academic research for accelerated computing tasks. As a result, even when AMD has robust hardware solutions, many developers still find it easier to build AI infrastructure around Nvidia products.

The significance of software has also impacted Intel, which has struggled to gain traction in the AI accelerator market. Intel’s Gaudi AI chips, which do not conform to typical GPU architecture, present additional challenges for the company. AMD is currently in a better position than Intel but continues to lag behind Nvidia.

Potential Changes with New AI Chips

Over time, the software landscape may improve, which could positively affect AMD's sales if their upcoming products resonate with the market. The MI325X is set for broad release by system manufacturers in the first quarter of this year, although it retains the same architecture as the earlier MI300X.

Looking ahead, AMD plans to introduce its next-generation Instinct MI350 family sometime in 2025, likely later in the year. AMD claims that the MI350 series will offer significant performance upgrades, featuring up to 35 times better AI inference performance compared to its predecessors. Following this, the MI400 family is expected to debut in 2026, based on a completely new architecture.

Despite ongoing product introductions, AMD faces fierce competition from Nvidia, and it will continue grappling with the challenge of its software disadvantage. Sales results for AMD’s AI chips this year remain uncertain; any distinct shortfall compared to expectations may lead to a decline in the stock price.

Timothy Green has positions in Intel. The Motley Fool has positions in and recommends Advanced Micro Devices, Intel, Nvidia, and Oracle. The Motley Fool recommends the following options: short February 2025 $27 calls on Intel. The Motley Fool has a disclosure policy.

AMD, AI, Sales