AMD: Markets are reducing the probability of a TAM breakout/expansion

Investment thesis

Advanced Micro Devices Company (Nasdaq: AMD) has lagged the overall pool of semiconductor stocks so far in 2024, performing in line with broader markets and offering a roughly 14% gain, marginally higher than the 12% year-to-date performance. of the S&P 500. In contrast, the iShares Semiconductor ETF (Sox) has returned 25% so far, while the VanEck Semiconductor ETF (Trait) posted gains nearly three times faster than AMD stock in 2024.

As I note in my analysis below, while AMD’s data center has shown some solid growth as we move through the initial post-pandemic phase of enterprise cloud infrastructure spending, the company has fallen from market share in the data center chip market with the success of OpenAI’s ChatGPT catching the world by surprise. Which led to a mad dash for Nvidia (NVDA) GPUs in their data center architecture.

The recent Computex 2024 event has reinforced my belief in the company’s growth, and I believe AMD is undervalued at current levels. I’ll begin my coverage of AMD by initiating an aggressive buy on the stock.

A quick summary of AMD’s revenue trajectory

AMD has successfully taken market share from Intel (INTC) over the past decade in the data center CPU market, which has focused on the x86 CPU architecture. As this decade went on, AMD also realigned its revenue segments to better segment its emerging revenue segments, such as data center revenue, as shown in Figure B.

In the past decade, AMD’s chips, such as the desktop-focused Ryzen series as well as its EPYC line of microprocessors intended for the server and embedded markets, have propelled the company forward as the company continues to win market share against Intel, as shown in the chart below In Figure C.

However, with the launch of ChatGPT in Q4FY22, the overnight popularity of OpenAI’s AI-powered chat assistant was enough to convince enterprises to upgrade their data center architectures to take advantage of the coming demand for accelerated machine learning workloads required by most ML models. The machine. Hyperscaler companies have led the way during these rapid capex expansion phases, in the last 15 to 18 months.

Unfortunately, companies like AMD and Intel have lagged behind in demand for chips that can efficiently handle accelerating cloud workloads related to machine learning. I think this was because the current crop of chips that AMD and Intel had on the market at the time were based on the x86 architecture and were no match for the throughput efficiency and large-scale data parallelism that Nvidia’s GPUs could provide at the same time. . Training large models. This was also shown in a rare missing earnings report in FY22, shown below.

Therefore, AMD, along with Intel, has lagged behind in the new market for AI-created accelerated computing chips, as shown in the chart below, with AMD and Intel’s market share falling further into the low single digits.

AMD is expected to gain market share in 2024

In Figure E above, you note that Nvidia’s GPU penetration was at its highest in 2023, but the forecast also estimates that AMD and Intel are expected to regain market share from Nvidia this year onward. While Intel’s market share is expected to be approximately 1%, AMD is expected to regain an additional 3% of the market.

There are two reasons I think this could happen with AMD. First, I’m encouraged by the launch of AMD chipsets in the last six to eight months, which puts the company in a better position. Last December, AMD launched its MI300 series APU, which aims to compete with Nvidia’s Hopper platform GPU lineup.

In one of the previous earnings calls, AMD management also surprised many analysts during the call by expanding TAM to $400 billion. Markets were expecting the TAM to be closer to the ~200B range. On that earnings call, AMD’s CEO added some additional details about why they expect a big upside.

I think we expect units to grow by a significant double-digit percentage. But you should also expect that the content will grow. So, if you think about how important memory and memory capacity will be as we move forward, you can imagine that we’ll see an acceleration there and the overall content just as we move to the more advanced technology nodes. So, there is some uptick in ASP there. And then, what we also do is we plan long-term roadmaps with our clients in terms of how they think about the size of training sets and the number of training sets. And then the fact that we think heuristics will actually overtake training as we get into the next couple of years with more organizations adopting.

I believe that in the next stage of AI, AMD chips will be more suitable for meeting the ML inference needs of enterprises with lower cost and higher chip efficiency. Below is a chart I’ve modified, which should illustrate how I see the next phase of AI evolving, in terms of model reasoning capabilities:

Additionally, at the latest Computex 2024, I was pleased to see more clarity from AMD’s CEO on the roadmap for accelerated AMD chips, which aligns closely with my vision for the next stage of AI. AMD will move to an annual release cycle for its accelerated GPUs this year and will eventually roll out its MI300 series to make way for the MI400 series in 2026.

I expect these strategic changes by management to be complementary to the market penetration I noted earlier in Figure C. I’ve also added AMD’s roadmap for accelerator chips below in Figure (g).

An example of the ramp can already be seen as management mentioned the demand for the MI3000 chipset launched last year. The company managed to achieve $1 billion in sales in two quarters. Building on the success, management raised its forecast guidance for MI300 chip sales in 2024 to $4 billion from the $3.5 billion previously expected. Note that the MI300 chip was just launched in December. Below is an excerpt from the latest BofA Securities 2024 GTC, held last week.

We launched the MI300X literally last December, right? We’ve assembled a $1 billion MI300X in less than two quarters. And when you think about it (the technical difficulty) and today, we talked about on the last earnings call, we have over 100 customers that we’re dealing with either in the development phase or in the deployment phase. So, we updated the number to over $4 billion in our latest earnings call.

AMD is up from current levels

Taking into account consensus forecasts, I expect AMD to increase its revenue in 2024 by at least double digits as the company becomes a beneficiary of the accelerating new chips it rolls out to its customers. Several companies have already announced their intentions to wean themselves off their reliance on Nvidia GPUs, and this could translate into a short-term tailwind for AMD’s product line. Additionally, the need for lower-cost modular inference chips will be a long-term tailwind for AMD. Therefore, I expect AMD to grow in line with consensus estimates, which puts the chip company at a CAGR of 19-20% over the next three years.

In terms of profitability, if I take the mid-range of consensus EPS estimates, I note that AMD’s EPS will eventually grow in strong double digits to FY26 EPS of $7.6. With AMD’s current market capitalization, this means that the markets currently value AMD at 23 times earnings in fiscal 2026.

Comparing that to the long-term growth rates of the S&P 500, the stock should trade at least on par with its 5-year average forward P/E of 43 based on its earnings growth rate, which translates to a share price of about $230.

This means there is at least a 40% upside from current levels.

Risks and other factors to consider

A slowdown in infrastructure spending poses the biggest risk to AMD at this point. If spending slows, AMD’s growth rates will suffer. One indicator I look for is expected capital expenditures by hyperscalers, who have previously indicated they will upgrade their data center architecture with updated chips and related infrastructure. I talked about this a couple of months ago in the Macro Conditions section of my SMH coverage.

Additionally, Apollo Global’s chief economist recently explained how continued strong earnings over the past few quarters indicate higher capital spending ahead, as shown in Figure J below. Monitoring these macro trends will be critical to formulating expectations for AMD as they relate to investors.

He stays away

There are enough reasons that I see that justify the value that AMD stock offers at current valuation levels. I’m encouraged by the additional detail and clarity in AMD’s accelerator chipset roadmap, which will position AMD to recapture market share in the data center market from 2024 onward, despite trailing Nvidia in second place. Additionally, changing trends in machine learning model inference needs will put AMD in a strong position to capitalize on future trends, and I expect AMD to reap the benefits as a result.

I highly recommend buying AMD at current levels.