AMD: The Attractive Valuation

1/ $Advanced Micro Devices(AMD)$ is at the edge of a bullish breakout.

Last time it bounced back from $90 levels, it went straight to all time highs with no catalyst.

This is repeating now and it's exploding data center revenue is a great catalyst.

Here is my AMD investment thesis: 🧵

Image

2.

You have to take a look at this:

Last time AMD bounced back off $90 levels, it tripled in a matter of months.

This was without any catalyst.

Now the same event seems to repeat with a catalyst: Data center revenue growth.

Image

3.

AMD data center revenue grew 69% last year.

From Q3 2023 to Q4 2024, quarterly revenue literally tripled increasing every quarter.

Yet, this is just the beginning.

AMD is finally catching up with $NVIDIA(NVDA)$ in performance.

Image

4.

MI300x now outperforms $NVIDIA(NVDA)$ H200 series chips in inference throughput by a margin for a friction of cost.

They achieved this by leveraging the SGLang framework, AMD’s new AI Tensor Engine for ROCm (AITER), and targeted hyper-parameter tuning.

Why is this important? Because inference is going to be a way bigger market than training.

Image

5.

Jensen Huang thinks inference workload will increase by billion times.

Training is where NVDA has an edge due to its widely adopted CUDA framework.

In inference, $Advanced Micro Devices(AMD)$ outperforms widely as we saw.

Thus, as the workload will shift to inference from training demand for AMD chips will explode.

6.

AMD total cost of ownership is also way lower than that of NVDA.

AMD MI300X sells approximately at quarter of the price for NVDA H100/H200 series chips.

As hyperscalers look for cutting costs as AI models mature, they'll increasingly turn to $AMD.

We saw this with META.

Image

7.

$Meta Platforms, Inc.(META)$ latest model now runs 100% on AMD chips for live traffic.

META VP of Fnfrastructure Supply Kevin Salvadori says AMD allowed them to make inference very efficiently.

8.

AMD is increasing its market share in data center chips.

NVDA market share in data center GPUs peaked in 2023 and started to decline.

As of 2024, Nvidia market share declined to 94% while AMD market share increased to 4.2%.

This may small but it's irrelevant for the stock performance as AMD is starting with way lower revenue base.

Any gain of market share will translate to substantial increase in revenue.

Image

9.

Valuation is also attractive.

AMD is currently trading at 23 times forward earnings.

It can easily grow revenues 25% annually in the next 5 years as demand for AI chips will only increase.

This gives us $85 billion revenue for FY 2030.

Assuming 30% net margin, we will have $25.5 billion net income.

Assuming a conservative 20 times earnings multiple, we will get a $510 billion company, more than double of today's valuation.

Given the quality of the company and low hurdle rate, I would buy this everyday.

Image