Advanced Micro Devices: Gaining Momentum in the Slipstream (AMD)

Advanced Micro Devices, Inc.Nasdaq:Am) recently reported weak growth figures in its first quarter earnings, which, combined with disappointing guidance, is making investors less confident about buying the stock right now. Additionally, the current balance of power is such that Nvidia and With Intel sharing the top spot in GPU and CPU related markets, a natural question is what AMD’s growth prospects are likely to reinvigorate market participants. In this article, we will explore this question in more detail as we examine the market potential and resulting valuation valuations for AI accelerators, gaming GPUs and commercial APUs.

The struggle to grow

As artificial intelligence continues to rapidly develop and expand into more and more sectors of the economy, AMD appears to be one of the key players to benefit. ChatGPT has given the company a new $1 billion TAM, but the price-performance chart below shows AMD falling further and further behind Nvidia Corp.NVDA).

After the release of its first-quarter earnings, investors don’t seem all that optimistic going into 2024. AMD $5.5 billion Quarterly revenue was down 11% sequentially. Data center grew 2% on a quarterly basis (up 80% year-over-year due to a low base effect), but that wasn’t enough to offset weaker performance in other segments. Additionally, overall guidance is somewhat lackluster, even though management expects data center revenue to grow 2.5% in 2024 versus 2025. Over $4 billion That’s up from a forecast of $3.5 billion in January.

Clearly, Nvidia has the upper hand in the data center AI accelerator space due to the high demand for its Hopper infrastructure and its technological leadership with its upcoming Blackwell GPUs. In gaming and visualization, NVDA also leads the market for discrete AI GPUs, and is ahead of Intel in the desktop and notebook CPU market.International Trade Commission) accounts for the majority.

Looking at valuation, things don’t look so rosy. AMD is currently trading at a forward P/E ratio of 47.5x, about 24% higher than the three-year median of 38.2x. We can see that the ratio breached the +1 standard deviation line (one standard deviation above the median) earlier in the year, then dropped back toward the median near the Q1 earnings release date, and is now recovering.

With this in mind, AMD doesn’t seem like a bargain unless it gets back on track for growth. Where will these growth prospects come from? Since this is the main focus of this article, let’s start with the data center market.

Nvidia is 90 and above According to various reports, the data center GPU market share is the second highest since 2015. Although AI technology is still in the early stages of development, it has already reached the stage where the inference part accounts for one-third of the market share. Approximately 40% The growth of the data center market was evident in NVIDIA’s recent earnings report. Investments in AI chips are already substantial, and to sustain these numbers, it’s clear that these investments need to start benefiting the company’s bottom line. Now the focus is shifting to AI chips. Making Money We use it to train our own models, not from AI (or inference). In my latest post on NVDA I wrote:

Training a model is an ongoing process that must be performed continuously as it becomes more sophisticated, upgraded, and bug-free, so the more data that is generated through inference, the more data the model has to process, requiring a more powerful GPU solution.

The point is, we believe inference will grow rapidly in the future and that this is an area where Nvidia can step in. In particular, the MI300X has an advantage over Hopper in memory-intensive tasks and delivers leadership inference performance according to AMD benchmarks. Spec Comparison It was revealed that the MI300X comes with 192GB of HBM3 memory and 5.3TB/s of memory bandwidth, while the upcoming H200 will come with 141GB of GPU memory and 4.8TB/s of bandwidth. In addition, AMD has yet to incorporate Instinct features, so there is room for improvement for Instinct. Latest HBM3e Memory chips used in the soon to be released H200.

Now, let’s talk numbers: According to BofA, the TAM for AI chips is an estimated $45 billion and could reach $180 billion by 2027, implying a compound annual growth rate of 38%. And assuming inference currently accounts for 40% of the market, I believe that could double over the next three years as the rapid adoption of generative AI brings the technology to other industries beyond cloud service providers. Rumor has it With the MI300X priced at less than half the price of the Hopper, the former could have a significant advantage in terms of TCO, which is especially important as enterprises optimize their costs. As a result, we believe AMD has a good chance of capturing around 10% of the datacenter GPU market by 2027. Currently 3%With this in mind, the company’s data center division is likely to grow faster than the market’s estimated 38% CAGR.

When it comes to gaming, the balance of power is roughly even. Still, AMD is catching up with Nvidia, Shape 19% AIB market share stands at $2 billion. From a consumer perspective, NVDA has a clear advantage in the top-tier segment due to the unmatched ray tracing performance of the RTX 4080/4090 series. In the budget segment, Nvidia has an advantage over AMD due to its superior DLSS upscaling technology. Even gamers on a budget can opt for the RTX 4050/4060 and enjoy the latest games at ultra-fast settings. And finally, we move to the mid-range, which is the most important in my view. Traditionally, gamers target this range because this GPU can deliver a high-quality experience at 4K resolution. Benchmarking the $699 Radeon RX 7900 XT card, we can say that this card can deliver a high-quality experience at 4K resolution. on average At 4K Ultra settings, it has better rasterization performance than Nvidia’s latest RTX 4070 Ti Super, which is $100 cheaper. Some would say that ray tracing makes RTX cards perform better, and they would be right. But with RT turned on, it averages less than 60 FPS. Some would argue that frame generation (FG) would ensure that level of frames. But FG technology allows you to add interpolation between frames, but Liquidity will not improve Because gamers can’t see those extra frames. As a result, the downside of this technology becomes clear: the gaming experience is marred by increased input latency. And when you start researching the gaming community, you find overwhelmingly negative feedback regarding the use of FG technology, with in-game characters looking “drunk”. No doubt, FG is useful in the budget segment, but not in mid-range GPUs. This gives AMD growth prospects and an opportunity to challenge Nvidia in this GPU price range that is important to gamers.

Finally, regarding the CPU market, after the extended refresh cycle due to COVID-19, 2024 is expected to be the peak year for the CPU market. chapter One As adoption of AI-enabled PCs increases, the replacement cycle for AI-enabled desktops and laptops will pick up, especially in the enterprise segment, which, combined with the adoption of Windows 11, could drive a recovery in PC shipments this year. Quite outdatedthis is, AMD’s new Ryzen Pro 8000 and 8040 series processors. Currently in the seed stage, the market is seeing a rise in AMD’s new APUs, with 54% faster Compared to Intel’s Core Ultra vPro lineup, it offers an average of 30% better CPU and 30% better graphics performance. The market expects 50 million AI-enabled PCs to ship in 2024, and I see this as a solid opportunity for growth for AMD at this point.

Risk factor

The main risk is the possibility of further tightening of monetary policy, which could limit companies’ budgets for generative AI investments. The recent rise in AMD stock price has been largely driven by ambitious forecasts for revenues in the data center segment. Investors may adjust their positions significantly if the market for AI graphics accelerators fails to meet growth expectations.

Investment Conclusion

Consensus EPS growth of 33% this year is expected, roughly in line with the expansion of its Instinct GPU and EPYC processor offerings. This lineup gives AMD leadership in inference and compute performance with a significant cost per token advantage.

Going forward, analysts are predicting a 58% increase in EPS in the next fiscal year. With such growth rates, I’m not worried about the P/E chart above, which shows AMD’s valuation in the high range. The forecast is reasonable from my perspective, as the AI-enabled hardware refresh cycle is likely to get into full swing next year. Plus, given management’s confidence in the demand for MI300X and the recovery of the embedded segment, I think the company can easily achieve such growth rates.

As a result, my price target for AMD is $263 per share, implying a ~50% upside from current market levels, implied by the $5.55 EPS forecast. Therefore, I recommend buying AMD shares and ignore the lackluster Q2 outlook. The inference business could offer a bigger growth opportunity than training over the next few years, especially as companies start to commercialize the potential of generative AI. This is where AMD could challenge Nvidia for less demanding memory-heavy computing resources. Moreover, the APU market for AI-enabled computers is developing, giving AMD ample opportunity to strengthen its growth potential.