AMD Q3 Preview: Strong Blackwell Competitor
Contributor since 2023 / 3115 Followers
Summary:
- Advanced Micro Devices, Inc. shares have surged 11% since August, outperforming the market, driven by investor confidence in their new MI325X AI chip ahead of Q3 earnings.
- The MI325X chip, with advanced AI capabilities, is set to challenge Nvidia’s Blackwell GPU by offering competitive performance at a lower total cost of ownership.
- AMD’s strategy includes enhancing their ROCm software framework to ease transitions from Nvidia’s CUDA, aiming to capture more market share in AI inferencing.
- Heading into earnings, AMD’s strong momentum and innovative chip lineup position them well to capitalize on the growing demand for AI inferencing solutions.
JHVEPhoto
Investment Thesis
Advanced Micro Devices, Inc. (NASDAQ:AMD) shares are now up 11.00% since my last write-up on the chip giant in August. This has outpaced the market’s 6.47% return in the same timeframe. I think investor confidence is riding high ahead of their earnings report on October 29. This confidence is well-supported in my opinion. The company came out with a game-changing series of chips.
This upcoming quarter is crucial since the company now sports a chip lineup that is highly competitive with the Nvidia (NVDA) Blackwell GPU series. AMD is now supported by their new AI chip, the Instinct MI325X, which directly competes with Nvidia’s Blackwell chip.
AMD chip’s new advanced AI capabilities now allows the company to provide stronger inference computing, which is critical for the future of AI workloads. Strong large language model, or LLM, performance is increasingly becoming dependent on model inference capabilities. We need a whole new lineup of chips to help power this.
The upside here is immense: if the new chip gets the nod of developers and cloud service providers, it will likely challenge Nvidia’s pricing power and market share.
AMD’s strategy includes improving their software framework, ROCm, to facilitate transitions for developers from Nvidia’s CUDA ecosystem to their own, to reduce the barriers to adoption. The company claims that the MI325X’s architecture promises to deliver up to 40% more inference performance on specific AI models compared to Nvidia’s offerings like the H200. Compared to Blackwell, the offering is strong as well.
Heading into earnings, I am still a strong buy on AMD. These advancements are real and put the chips well in line with its Blackwell competitor.
Why I’m Doing Follow-Up Coverage
AMD shares have outperformed the broader market since my last coverage, as I think the market increasingly recognizes their advancements in chip technology through the MI325X AI chip. I truly believe it’s a viable alternative to Nvidia’s Blackwell version.
Since my previous analysis, AMD has also benefited from the interest in advanced reasoning AI models, notably OpenAI’s new o1 model and Anthropic’s Claude 3.5. These have integrated advanced reasoning capabilities that are crucial for applications like programming and scientific research to help enhance real-time problem-solving abilities instead of merely processing vast datasets. This is the next leg up in AI technology, powered by new inferencing architectures.
AMD’s advantage lies in their chip architecture that, I think, are well suited for inference. The industry is honing in on more efficient and affordable GPUs for inferencing. This supports the capabilities of models like o1 and Claude 3.5.
The purpose of my follow-up coverage is to show how these developments create a specific opportunity for AMD to differentiate themselves. Inferencing is like the wild west for AI right now. AMD is a well-suited cowboy up for the challenge.
Why They Are A Strong Blackwell Competitor
AMD’s new MI325X GPU is set to enter the market this quarter (Q4 2024) to help the California-based company capitalize on the growing demand for AI inferencing needs.
The MI325X has a new memory architecture, which is designed to accelerate AI calculations to better meet the demands of high-performance AI applications that require speed and efficiency.
In addition, the MI325X employs AMD’s new 2.5D chip architecture, a result of their acquisition of Xilinx four years ago. This architecture allows for increased performance and efficiency that in plenty of ways Nvidia can’t copy (because they don’t use 2.5/3D chip technology).
While some reports indicate that the MI325X chip will not beat the Blackwell chip on all fronts (in terms of performance) the chip beats the current H200 system delivering 40% more inferencing performance.
Here’s where I think it becomes competitive to Blackwell: AMD has committed to lowering the total cost of ownership, which makes the MI325X appealing to customers. AMD plans to undercut Nvidia’s pricing strategy. The Blackwell GPU is set to cost between $30,000 to $40,000 per GPU. AMD’s current MI300X costs $10,000-$20,000 per GPU. The new GPU should come in a similar price range.
Inferencing is the future of AI compute. While Blackwell may put out stronger inferencing capabilities, this new GPU means we can get much faster inferencing for a lower cost.
Put it this way: new models like the o1 currently run on the H200 GPU from Nvidia. This chip is now 40% faster than the H200 which means many of the key tasks o1 could be used for (but right now, it’s too slow) become far more feasible once it’s run on a MI325X chip. This means affordable inferencing that is at a speed that makes models usable is just around the corner. That’s powerful.
Q3 Preview
AMD is set to announce their Q3 earnings after the market October 29. Currently, expectations for EPS are set at $0.92, putting year-over-year growth at a strong 30.83%.
Revenue expectations for Q3 are currently pegged at $6.71 billion, or a 15.77% year-over-year uptick. I actually think this is a conservative estimate given how strong the MI300X chip has been selling.
Wall Street analysts are starting to price in more strong momentum, especially given the company’s performance in the data center space. Analysts estimate quarter-over-quarter data center revenue to jump 21%, and come in with year-over-year growth at an impressive 114%.
During the upcoming earnings call, I’ll be looking for explicit data on how AMD’s new chips can compete against Nvidia’s Blackwell. I expect the company executives to focus their earnings call on both the MI300 and MI325X chips. The market is clearly big. I am excited to see more evidence showing they are capturing more of it.
Valuation
AMD shares are definitely expensive on the surface, reflecting the high expectations the chip company has from investors. With a non-GAAP forward P/E ratio of 45.27, this well exceeds the sector median of 24.19. This is not where I want to focus this analysis.
However, when we dive into the forward PEG ratio, which stands at 1.12, this suggests the market is valuing shares discounted to where they should be given the earnings growth we are supposed to see over the next 12 months. The sector median forward PEG is 1.86. In essence, the market is implying that earnings will not be as strong as expected, which means their real forward PEG is well above this 1.12 figure. I disagree. I think EPS estimates are conservative, not optimistic.
If AMD’s shares were to converge on this sector median forward PEG ratio of 1.86, this would imply a 66.07% upside in shares as long as the company can meet or exceed current growth expectations.
Risks
Let’s be clear: AMD (for most of the last 10 years) has been directly competing with Nvidia. There’s no doubt that competition remains intense, now driven by their Blackwell series. Really, the market’s fair value of AMD rises and falls based on how competitive the market thinks AMD’s chips are compared to Nvidia.
Nvidia currently holds a commanding position in the AI chip market, with an estimated 94% market share. Don’t get me wrong, these new Blackwell chips are going to make it hard for AMD to crack into this space and grab more market share. But I think the MI325X will make a strong entrance.
I believe there is a growing sense of optimism regarding AMD’s capability to succeed. Inferencing is still up for grabs (as a market). Most of Nvidia’s market share is in the training side of the AI GPU market (not the inferencing side) These new chips are key for affordable inference. Cost is starting to become a big factor given the wide availability of different AI models.
Speaking on their competition, CEO Su said in an interview:
…probably the biggest thing that sets us apart in our strategy is that we really believe in end-to-end AI in every aspect. There are companies that are working on some aspect of AI. Our view is, hey, AI is going to be everywhere. AI is going to be throughout our entire product portfolio. Whether you’re talking about, in your clients—so in your PCs, or things that that you use personally, or if you go into a retail store, they’re going to have AI there locally so that it makes that service delivery more efficient—or in the cloud, where you’re training the largest foundational models, or you’re doing your ChatGPT queries, and so on. Our view is AI is end-to-end in every aspect of our portfolio.
The company plans to be an end-to-end AI chip solution provider. I think this is a winning recipe.
Bottom Line
Heading into earnings, AMD (and its shares) has strong momentum, driven by investor confidence in their newly launched Instinct MI325X AI chip. This upcoming quarter is critical, but I think the chipmaker is going to rise to the challenge.
The company has a unique opportunity to capitalize on this transition to inference compute in AI.
The MI325X is expected to help capture market share. Demand here is growing, but has not yet been fully saturated. I think we will see in the upcoming earnings report on October 29 that AMD is starting to convert this potential into tangible market share.
Given the high stakes, the success of the MI325X will be pivotal in keeping AMD at pace with the leadership of Nvidia. I think it’s combination of inferencing strength and price will keep them there. With this, I am still a strong buy on shares.
Analyst’s Disclosure: I/we have a beneficial long position in the shares of AMD either through stock ownership, options, or other derivatives. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Noah Cox (main account author) is the managing partner of Noah’s Arc Capital Management. His views in this article are not necessarily reflective of the firms. Nothing contained in this note is intended as investment advice. It is solely for informational purposes. Invest at your own risk.
Seeking Alpha’s Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.