AMD's AI Race: Patience Might Just Pay Off

Large Caps 0 replies 0 votes 1473 views

In May 2023, we took a closer look at Advanced Micro Devices, Inc. (NASDAQ: AMD) and dubbed it a Buy, citing the surging demand for generative AI as a potential catalyst for the semiconductor industry's next super-cycle. With the strategic leadership of CEO Lisa Su and Victor Peng (formerly of Xilinx), we believed that AMD could make strides towards closing the gap with Nvidia (NVDA), thereby bolstering its financial performance.

AMD's Q2 2023 Performance

However, in light of Nvidia's stellar performance in the second quarter of 2024, boasting data center sales of $10.32 billion (a staggering 141.1% increase QoQ and 171.5% YoY), it's evident that AMD investors might be feeling somewhat left behind. In the same period, AMD reported underwhelming data center revenues of $1.32 billion (a mere 2.3% uptick QoQ but a 10.8% decline YoY), echoing the sentiment expressed by numerous market analysts.

Most concerning, perhaps, is the drastic decline in AMD's data center segment's operating margin, down to 11.1% (a 0.3 point decrease QoQ and a significant 20.6% decrease YoY). Furthermore, AMD's overall inventories surged to $4.56 billion, a 7.8% QoQ increase and a staggering 72.7% YoY jump.

Even the latest EPYC Bergamo CPU, despite its touted market-leading performance for cloud optimization, failed to meet consumer demand expectations, prompting AMD to slash MSRPs by up to 20%. This starkly contrasts with Nvidia's expanding gross margins, at 70.9% (a 6.3 point increase QoQ and 25.6% YoY), and operating margins at 51.4% (a substantial 21.6 point increase QoQ and 41.3% YoY). Clearly, Nvidia is enjoying robust consumer demand, despite its premium pricing.

AMD's Outlook in Q3 2023

Unfortunately, AMD's Q3 2023 revenue guidance of $5.7 billion (a 6.5% QoQ increase and a modest 2.5% YoY uptick) doesn't inspire confidence. This is mainly attributed to the slower sampling of the MI300 in Q3 2023 and its production ramp, expected only in Q4 2023.

The big question is whether AMD's MI300 Chip Accelerator can effectively chip away at Nvidia's market share, especially considering Nvidia's bold projection of Q3 2023 revenues reaching $16 billion (an 18.4% QoQ increase and an astounding 169.8% YoY surge).

Given the assumed cadence from Q2 2023, we could extrapolate a hefty data center weighting of $12.2 billion in Q3 2023 (an 18.2% QoQ increase and a remarkable 218.5% YoY surge). This implies another substantial GPU win for Nvidia's Jensen Huang, and, to a lesser extent, for Intel (INTC), since the DGX H100 utilizes the Sapphire Rapids CPU.

Nvidia's remarkable growth is largely attributed to "accelerating demand from cloud service providers and large consumer Internet companies for the HGX platform, the engine of generative AI and large language models, based on our Hopper and Ampere architecture Tensor Core GPUs."

However, uncertainties surrounding compatibility might hinder providers and consumers from diversifying their AI platforms with AMD's MI300. Nvidia's aggressive launch of the next-gen GH200 Grace-Hopper superchip, a direct competitor to AMD's MI300, set for delivery by Q2 2024, could exacerbate this challenge.

In light of these developments, while Lisa Su reported a more than sevenfold increase in "AI cluster engagements" sequentially, the successful conversion of these engagements into orders and sales remains uncertain.

This uncertainty is compounded by Nvidia's commitment to incremental HGX supply every quarter through 2024, supported by Taiwan Semiconductor Manufacturing Company Limited's (TSM) expanded CoWoS advanced packaging capacity for AI chips, which is set to increase from the original monthly capacity of 8,000 wafers in H1 2023 to 9,000 in H2 2023 and around 25,000 in 2024.

With AMD securing only a monthly capacity of 6,000 wafers in Q4 2023, it appears that the Lisa Su-led company might face some initial setbacks in the AI race. As a result, investors should manage their short-term expectations accordingly.

So, Should You Invest in AMD?

As of now, AMD is trading at a next-twelve-month enterprise value-to-revenue (NTM EV/Revenue) ratio of 6.97x and a next-twelve-month price-to-earnings (NTM P/E) ratio of 30.26x. These figures are higher than its one-year means of 5.67x and 26.71x, respectively, although the P/E ratio remains relatively moderate compared to its pre-pandemic mean of 4.31x and 36.99x, respectively.

It's important to note that while AMD has historically traded closely with Nvidia in terms of valuations, there has been a noticeable divergence since May 2021. Nvidia now leads with a NTM P/E of 43.90x, thanks to its diverse chip and Software-as-a-Service (SaaS) offerings across data centers, AI, IoT, gaming, and automotive end-markets.

Nevertheless, AMD still maintains a significant premium compared to the semiconductor median P/E valuation of 19.32x. This suggests that it hasn't been entirely excluded from the market's enthusiasm for generative AI.

Currently, AMD's stock has retraced much of its gains from June 2023 and is now testing support levels from August 2022 and March 2023, around the $100 mark. The rapid rise and subsequent correction in the generative AI sector have affected many related stocks, such as Super Micro Computer, Inc. (SMCI) and C3.ai (AI), which have also retraced over the past few weeks.

However, this correction presents an opportunity for savvy investors. It has improved the upside potential for our long-term price target of $160, based on AMD's NTM P/E of 30.26x and the consensus FY2025 adjusted EPS of $5.29.

In conclusion, we maintain our belief that although Nvidia may have claimed victory in the initial round of the AI race, AMD has the potential to share the spotlight in the long run, depending on the timing of its product launches. Lisa Su's hint at the next-gen MI400 offerings, along with a tailored lineup for the Chinese market, akin to Nvidia's H800 and A800 GPUs, demonstrates the company's strategic approach.

Combined with promising reports of the MI300 potentially outperforming Nvidia's H100, we believe the Total Addressable Market for AI chips, estimated to be $150 billion by 2028, is ample to accommodate multiple winners.

Given the attractive risk/reward ratio, we continue to rate AMD stock as a Buy for patient investors. The AI race is far from over, and AMD's journey is one worth watching closely.

Forgot Password?
Don't have an account? Sign up