Summary

• AMD has declined 45% from its highs in early 2024.

• Nvidia leads in AI chip performance and ecosystem, but AMD offers better value for money.

• Considering the market growth rates of AMD's underlying business segments, the company could be undervalued by up to 34%.

AMD declined significantly in the last months and is currently down 45% from their highs in March 2024.

Data by YCharts

AMD's biggest competitor Nvidia (NVDA) also declined notably by 22% from the highs in January. They however already bounced back 12% to only -10%. This was mainly caused by DeepSeek and the uncertainties caused by potential trade tariffs from the Trump administration, that could negatively impact both AMD and Nvidia. AMD is generating ~23% of the revenue in China while NVDA is currently at ~14%.

Data by YCharts

In my latest Nvidia analysis (4th Feb.) I rated the company a Strong Buy considering the compelling growth prospects as well as the recent decline.

In the article, I am going to compare the two companies to see if AMD is offering similar or even better potential.

Who is leading the AI race?

Providing chips for AI applications is the critical business segment for both companies. Nvidia is currently generating 88% of their revenue in the data center GPU segment while AMD is doing 50%.

In AMD's latest earnings the company achieved a record revenue of $7.7 billion in Q4, a YoY increase of 24%. This increase was mainly caused by the strong growth rates of 69% YoY and 58% YoY in the data center and client segment.

For Q1 of 2025 AMD guided revenues of $7.1 billion, which beats analyst expectations. A QoQ decline from Q4 to Q1 seems to be normal for AMD.

AMD's Quarterly Revenue (seekingalpha.com)

The CEO, Lisa Su, however, also anticipated a 7% decline in the data center revenue. Furthermore, they won't be giving any forecasts on AI chip revenues anymore, which led to unrest among investors.

Their production of the new high performance MI350 series is planned to start mid-2025 and is backed by strong customer demand. According to Su, this new generation is a 35x improvement compared to existing products. She also reiterated the confidence that the whole AI portion of the business should grow by several billions of revenue in the next years.

Meta (META) was using the MI300X chips to power their Meta AI Llama 40B model and Microsoft (MSFT) used this chip generation for GPT-4 copilot services.

Looking at the benchmark of the products, Nvidia's H100 chips seems to be ~7% better than the MI300X on the Llama 70B inference. Assuming recent GPU prices, this means that AMD is offering a litter more "bang for the buck".

AMD vs. Nvidia Performance Benchmark (MLPerf Benchmark)

Comparing AMD Radeon 7900 XTX to Nvidia's GeForce RTX 4090 and 5090, we get to a similar picture. This was recently published by Nvidia to show the differences in the state of the art chips when using DeepSeek.

AMD vs. Nvidia Performance Benchmark (nvidia.com)

The RTX 4090 is outperforming the 7900 XTX by 46%, but customers have to pay up to 89% more. When customers decide to spend 2,300 Euro or 156% more for the RTX 5090, they can achieve performance improvements of up to 103%. The RTX 4090 is currently the go-to GPU for deep-learning and AI applications.

Nvidia's software ecosystem, especially CUDA (Compute Unified Device Architecture), gives them the upper hand in the AI and data center businesses rather than just its potent hardware. It has been difficult for rivals like AMD, Intel (INTC) and Qualcomm (QCOM) to replicate the success.

Due to the software, networking and infrastructure synergies, Nvidia is creating full-stack solutions that maximize the performance. These managed to improve the initial A100 chips - introduced 4.5 years ago - by 12x. This is around 3 times better than Moore's Law would suggest, further emphasizing the importance of Nvidia's full-stack solutions for AI and the importance of the infrastructure surrounding high-performance chips.

To summarize, AMD's chips are less performant than the competing Nvidia chips. They, however, tend to be the better value, providing more computing power for less investment. Especially Nvidia's full-stack solutions led to higher performances.

While I certainly believe that there is enough room in the data center GPU market for both companies, I think that Nvidia's position seems to be more favorable. Should the pressure from AMD increase they could - as of now - just reduce their prices of the high-quality chips to match AMD's less performant ones.

Keeping this in mind, AMD could, on the first look, still be an attractive investment considering the decline of ~45% over the last months and the lower PE ratio.

Data by YCharts

Valuation

To evaluate the company independently of any peers, I used a Discounted Cash Flow Analysis that calculated the required growth rates to justify the current valuation. After that, we will analyze the current business segments and the underlying market growth prospects to see if these growth rates are achievable for AMD.

For the EBIT margin in the DCF, I assumed the following improvement over the next years, as the company is increasingly improving their margins.

Chart

Their historical EBIT to Free Cash Flow conversion is very high (2022: 307%, 2023: 576%, 2024: 168%). This should be mainly caused by deferred revenue, which is why I assumed an optimistic but more realistic long-term conversion rate of 100%.

Required Growth

AMD Discounted Cash Flow Analysis: Required Growth (I) (own assumptions)

AMD Discounted Cash Flow Analysis: Required Growth (II) (own assumptions)

With these assumptions, AMD needs to achieve top-line growth rates of 23% p.a. for the next eight years to justify the current valuation.

Realistic Growth Rates

The company is mainly offering Graphic Processing Units (GPU) and Central Processing Units (CPU) for Data Centers and desktop and notebook personal computers.

Experts predict the global GPU market to grow by annual rates of 28.6% to 33.2% in the next years. For the CPU segment, experts anticipate growth of 15.2% p.a. in the same time-frame. As AMD is increasingly focusing on developing AI GPU's for the data center and client segment, I think it is fair to assume that these growth rates can be weighted by 80/20 to be applied to AMD's growth. This results in a realistic growth rate of 27.8% p.a. for AMD (also assuming the average of the two GPU growth rates).

Assuming this rate and keeping the other metrics identical to the first DCF, we get the following valuation.

AMD Discounted Cash Flow Analysis: Realistic Growth (I) (own assumptions)

AMD Discounted Cash Flow Analysis: Realistic Growth (II) (own assumptions)

Considering the assumptions, we get a fair value market cap of 234 billion, indicating that the company could currently be undervalued by 34%.

Risks To Consider

Like calculated, the market is expecting growth of 23%+ each year. Should the growth slow or other hiccups happen, this can lead to multiple contractions and sharp declines.

Capacity and demand in the semiconductor market historically behaved cyclical as these two aspects were following these steps:

1. Demand Increase -> Capacity Shortfall

2. Demand Increase/Constant -> Capacity Increase

3. Demand Constant -> Capacity Surplus

4. Demand Constant -> Capacity Constant/Decrease

5. back to 1.

Global Semiconductor Growth In Capacity (Chart of the Day)

We are currently at step two. Going beyond this step and causing a capacity surplus, prices and profitability margins will decline for AMD.

Conclusion

Despite Nvidia's dominant market position, I am currently rating AMD a Buy, considering the potential upside of 34% when anticipating optimistic yet realistic growth rates.

Investors that are looking for additional exposure in the semiconductor market, could use these price levels to enter a position. If I had to choose between Nvidia and AMD, I would currently pick the first one as NVDA could still be undervalued by 28% - more on that here - and because of NVDA's position as the technology leader.