AMD is about to update investors on its artificial intelligence chips.
Nvidia (NVDA 4.14%) is the leading supplier of graphics processing chips (GPUs) for the data center, which developers use to train and inference artificial intelligence (AI) models. The company’s CEO, Jensen Huang, believes data center operators will spend $1 trillion building AI infrastructure over the next five years, so it’s an incredibly valuable opportunity.
Nvidia had an estimated 98% share in the market for data center GPUs last year, but competition is ramping up, and Advanced Micro Devices (AMD 1.24%) might be one of the company’s biggest threats.
AMD is becoming a multifaceted AI chip company
AMD’s chips power some of the world’s most popular consumer electronics, from Sony‘s PlayStation 5 to the infotainment systems in Tesla‘s electric vehicles. AMD’s GPUs, CPUs (central processing units), and NPUs (neural processing units) are also extremely popular with manufacturers of personal computers.
In fact, the company’s latest Ryzen AI 300 series NPUs are set to power more than 100 computing platforms from manufacturers like Microsoft, Asus, Lenovo, and HP. Ryzen AI chips were designed to process AI workloads on-device, which reduces the reliance on external data centers to create a faster user experience.
But as mentioned at the top, AMD is quickly catching up to Nvidia in the market for data center GPUs. AMD is scheduled to report its financial results for the third quarter of 2024 (ended Sept. 30) on Oct. 29. Here are two numbers related to that segment that investors need to watch.
Image source: Getty Images.
The first number to watch: Data center revenue
AMD started shipping its MI300X data center GPU, which competes directly with Nvidia’s flagship H100, at the end of last year. It has attracted a few of Nvidia’s biggest customers already, including Microsoft, Oracle, and Meta Platforms, and some of them are yielding performance and cost advantages by using the MI300X compared to the H100.
The MI300 series generated $1 billion in sales within two quarters of its launch, making it the fastest-ramping product in AMD’s history. Naturally, AI GPUs will be the driving force behind the company’s data center revenue going forward.
During the first quarter of 2024 (ended March 30), AMD generated $2.3 billion in data center revenue, which was an 80% increase from the year-ago period. Then, during the second quarter (ended June 29), the company delivered $2.8 billion in data center sales, a whopping 114% year-over-year jump.
Simply put, AMD’s data center revenue is accelerating, and investors should look for that trend to continue in Q3. If it does, it would be a very positive sign that the MI300 is taking market share. Since the data center segment now accounts for almost half of the company’s total revenue, it’s going to be the center of attention in the upcoming Q3 report.
The second number to watch: Data center GPU sales guidance
The MI300X was just the beginning of AMD’s foray into AI data center GPUs. At the company’s “Advancing AI” event earlier this month, CEO Lisa Su offered fresh details about the upcoming MI325X and the game-changing MI350X.
The MI325X GPU will start shipping in the first quarter of 2025, and it will compete directly with Nvidia’s H200 GPU (the successor to the H100). Su says it has 80% more high-bandwidth memory and delivers 30% better inference performance than the H200. Therefore, even though AMD’s chips are launching several months behind Nvidia’s, the company is delivering a technical advantage that could help attract customers.
However, AMD will soon be on the back foot yet again, because Nvidia is scheduled to start shipping GPUs built on its new Blackwell architecture during its fiscal 2025 fourth quarter (which runs from November to January). Blackwell-based GPU systems like the GB200 offer up to 30 times the inference performance compared to the equivalent H100 systems, and they will provide a substantial improvement in cost efficiency.
AMD is already working on a GPU architecture to rival Blackwell, called CNDA (Compute DNA) 4. It’s the foundation for the upcoming MI350X GPU, which will offer up to 35 times greater performance than CDNA 3 chips like the MI300. The MI350X won’t start shipping until the second half of 2025, but keep an eye out for any updates from the company during its Q3 earnings call with investors.
That brings me to AMD’s GPU sales forecast. The company came into 2024 expecting to sell $3.5 billion worth of data center GPUs for the year, but it revised that estimate higher in both the first and second quarters, and it now stands at $4.5 billion. Another upward revision would suggest demand is still accelerating for the MI300, which would be further evidence that a growing list of customers view AMD as a true alternative to Nvidia.
Plus, accelerating demand for the MI300 bodes very well for MI325X and MI350X sales in 2025, and since the stock market is a forward-looking machine, that would be great news for AMD shares heading into the new year.
Randi Zuckerberg, a former director of market development and spokeswoman for Facebook and sister to Meta Platforms CEO Mark Zuckerberg, is a member of The Motley Fool’s board of directors. Anthony Di Pizio has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, HP, Meta Platforms, Microsoft, Nvidia, Oracle, and Tesla. The Motley Fool recommends the following options: long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy.
