AMD Ahead Of Earnings – A Contrarian View

Ahead of earnings, there is little doubt that the sentiment in Advanced Micro Devices (AMD) stock is very poor. The stock is down nearly 50% from its 2024 peak and recent analyst commentary and option activity suggests that investors expect further downside from the Q4 earnings call. The biggest reason for the negative sentiment is that several analysts and funds are now expecting that 2025 Instinct sales will be in the $7B+ range – i.e. virtually no growth or negative growth from the expected Q4 2024 levels. As a result, the Q1 and full year 2025 estimates for the Company, including for the high visibility Instinct accelerator sales, have been steadily trending down going into Q4 earnings.

The Bear Case

The reason for pessimism seems to be multi-fold:

1. Supply chain centric analysts are showcasing relatively soft demand for MI325. There are widespread concerns about a) AMD not growing its presence outside of Meta (META) and Microsoft (MSFT) and b) AMD not maintaining GPU share at both of these customers. Most analysts have taken a view that MI325 revenues will stall or decline as Nvidia (NVDA) Blackwell ramps and expect that the upcoming MI355 will not meaningfully change the demand profile. Some expect MI355 launch to be late in the year or otherwise to have no meaningful contribution in 2025. Some are waiting for MI355 to be launched before assessing its merit with a concern that Blackwell Ultra may overshadow it even before launch. All in all, very few see AMD gaining competitively with MI355 given Blackwell and Blackwell Ultra.

2. There are concerns that AMD may also underperform in the PC and traditional server markets due to weak demand. Many expect the weakness to be further exacerbated by demand being pulled into Q4 ahead of any new Trump administration tariffs - potentially making Q1 softer than typical seasonality.

3. Intel (INTC) has made the sentiment dourer by essentially promising a price war to retain market share.

Adding to the above, there is a perennial bunch of misguided analysts who expect that either Intel or Nvidia to take market share from AMD in CPUs or GPUs. That AMD has been steadily gaining on Intel and Nvidia seems to be lost on this crop of analysts.

As a result of these concerns, expectations for Q1 and 2025 guidance have been sliding steadily over the last several weeks (image below)

Note that the Q4 2024 expectations have held steady since last earnings report and the downward revisions have to do with 2025 and 2026 numbers. In the short term, Q1 revenues are expected to drop to about $7B compared to $7.5B in Q4.

It is not that there is no merit to the bear case. Outside of supply chain data points, evidence is accumulating that pre-training scaling limitations are changing the AI demand profile. There is now a looming risk that training demand will reduce dramatically from previous rosy expectations. Nvidia 2025 expectations have moved down meaningfully over the last few months. While many see the lowered expectations as an outcome of Blackwell delays, this is not an accurate assessment. Had the overall demand not dropped, the lost Blackwell orders should have manifested as new Hopper orders. But that has been only partly true and the overall 2025 demand has declined. The optimistic forecasts for Nvidia seem to have dropped by as much as $50B for 2025. In context, a few billion-dollar downtick in AMD Instinct orders would not seem too unusual.

On the PC and server front, superficially the concerns have merit for analysts and investors who follow high level industry trends but lack appreciation for product level trends (that is most of the analyst and investment community).

The Bull Case

As much as the Bear case seems to have merit, the gloom and doom that drove down the stock is entirely misguided. Supply chain data and macro factors seem to be the main culprit here.

In Beyond The Hype’s view, semiconductor investing is more about products and applications. Supply chain data, as alluring as it may be, can be entirely misleading for various reasons. The main problem with supply chain data is that it is backward looking and can change at a drop of the hat.