Nvidia is the go-to supplier of graphics processing units (GPUs) for processing artificial intelligence (AI) workloads in the data center. Sales are soaring, with demand coming from some of the world's biggest tech companies like Microsoft and Amazon.
But Advanced Micro Devices (AMD +6.69%) has emerged as a worthy competitor, and its new data center GPUs are attracting some of Nvidia's most prized customers. However, AMD also makes the most powerful chips for AI-enabled personal computers, and it has taken a dominant position in that market.
AMD stock is trading 29% below its all-time high, but its recent financial report for the second quarter of 2024 (ended June 30) suggests investors might want to buy it right now.
Image source: Getty Images.
AMD is quickly becoming a leader in AI chips
Nvidia's H100 data center GPU set the standard for AI development, and AMD launched its MI300 lineup last year in an attempt to compete. The MI300X is a pure GPU, whereas the MI300A combines GPU and CPU hardware to create an accelerated processing unit (APU).
The response has been very positive, with customers finding both performance and cost advantages in the MI300X compared to the H100. During the second quarter, MI300 sales topped $1 billion for the first time, and AMD increased its full-year forecast to $4.5 billion (from $4 billion just three months ago). So far, customers include Microsoft, Oracle, and Meta Platforms, to name just a few -- all of which also buy chips from Nvidia.
But AMD can't rest for a single minute, because Nvidia is gearing up to launch next-generation chips based on its new Blackwell architecture this year, which will comfortably leapfrog the MI300 in terms of performance. As a result, AMD already plans to release the MI350 series in 2025. It will be built on its new Compute DNA (CDNA) 4 architecture, and it could deliver a whopping 35 times more performance than previous chips based on CDNA 3 (like the MI300). The MI350 will be in direct competition with Nvidia's Blackwell-based chips.
But the data center is just one piece of AMD's AI strategy. The company's Ryzen AI chips allow personal computers to process AI workloads on-device, which means chatbots and virtual assistants can be embedded into more applications. Since queries won't have to travel through an external data center, it creates a much faster user experience.
Earlier this year, AMD said millions of PCs had already shipped with its Ryzen AI chips, giving the company a market share of 90%. But it's building on that dominance with the new Ryzen AI 300 series for notebook computers, which is now the industry's fastest neural processing unit (NPU). The first notebooks went on sale last weekend, but AMD says more than 100 platforms are set to launch with Ryzen AI 300 series chips in the next few quarters, from leading notebook manufacturers including HP Inc., Acer, Asus, Lenovo, and more.
Revenue is soaring in two key segments thanks to AI
AMD generated $5.8 billion in total revenue during Q2, which was an increase of just 9% from the year-ago period. However, the real story lies beneath the surface.
AMD's data center revenue soared 115% year over year to a record $2.8 billion, thanks to the MI300. The company's client segment, which is home to its Ryzen AI computer chips, generated $1.5 billion in revenue, a 49% increase.
Despite those points of incredible strength, growth in AMD's headline revenue number was weighed down by continued weakness in its gaming and embedded businesses. Gaming delivered a revenue decline of 59% compared to the year-ago period, as demand for consoles like the Sony PlayStation 5 tapered off. The embedded segment, which is home to many chips used in industrial applications, saw revenue fall 41% due to high levels of existing inventory.
But investors should remain focused on the data center and client segments, which together make up 74% of AMD's total revenue. That's where most of the explosive AI growth is likely to come from for the foreseeable future.
NASDAQ: AMD
Key Data Points
Why AMD stock is a buy on the dip
As I mentioned earlier, AMD just raised its 2024 forecast for data center GPU sales to $4.5 billion. However, that's a drop in the bucket relative to the company's addressable opportunity. Nvidia CEO Jensen Huang predicts data center operators will spend a whopping $1 trillion over the next five years to meet demand from AI developers.
AMD is just breaking into the AI data center market with its MI300, but as it catches up to Nvidia's new Blackwell chips next year with the MI350, it will likely snatch more market share. Even capturing 10% of that $1 trillion figure will drive incredible growth in AMD's GPU sales.
With that said, AMD stock isn't cheap right now despite the 29% decline from its all-time high. Based on its trailing-12-month non-GAAP (adjusted) earnings of $2.78, it trades at a price-to-earnings (P/E) ratio of 52.5. That's a substantial premium to the Nasdaq-100 index, which trades at a P/E ratio of 30.9, implying AMD stock is overvalued.
However, the stock appears much cheaper when measured against its future earnings. Wall Street predicts AMD will generate $5.51 in earnings per share in 2025, placing the stock at a forward P/E ratio of just 26.4. In other words, investors who buy the stock today and hold on for at least the next two years could yield a very nice return.
With a market cap of just $236 billion, AMD is worth less than one-tenth of Nvidia's $2.8 trillion valuation. Considering the quality of its chips and the fact it's selling to many of Nvidia's customers, AMD stock has the potential to deliver more upside than its rival from here.
