Advanced Micro Devices (NASDAQ:AMD) stock has tumbled 36% from its 52-week high of $211 in March and trades around $135 as of writing. The decline has been driven by disappointing earnings and guidance that fell short of Wall Street expectations. This dip has some investors wondering if now is the time to buy AMD on the cheap. However, a closer look at AMD’s competitive position and growth prospects compared to rival Nvidia (NASDAQ:NVDA) suggests the choice may not be so clear-cut.
Before we get into why AMD stock declined, let’s first take a glance at its Q3 results.
However, the Q4 guidance disappointed analysts:
-
Forecasted Q4 revenues between $7.2 billion and $7.8 billion, below the consensus estimate of $7.55 billion.
-
This guidance miss caused a 13% stock drop over two trading days.
These headwinds have soured market sentiment on AMD in the near term despite the company still growing revenue by 18% year-over-year in Q3. The question is whether the 36% pullback represents a buying opportunity for a quality company facing temporary issues or a sign of tougher times ahead. But first, let’s look at what kind of competition AMD is facing.
Nvidia is AMD’s biggest competitor—we can all agree on that. That said, there are some differing opinions about how both companies could end up in the long run.
Nvidia is focusing more on high-margin AI chips, whereas AMD is trying to undercut it and offer more value. I personally see it as something copied from AMD’s playbook in the CPU market—where AMD slowly and methodically caught up to Intel—but it never managed to dominate it.
Nvidia currently has an edge over AMD in the high-end GPU market. It has AI features like DLSS but, AMD offers very competitive performance for the price with its RX 7000 series, especially in the mid-range.
Nvidia holds a clear lead with its GeForce RTX series—particularly the RTX 4090—in terms of raw gaming power and ray tracing capabilities. AMD’s Ryzen processors are highly competitive but do not challenge Nvidia’s dominance in the GPU space.
But again, this doesn’t really matter much. Analysts care much more about dedicated AI products than anything related to gaming.
Nvidia also leads here with its H200 Tensor Core GPUs, which are widely adopted for large-scale AI training tasks. AMD’s Instinct MI300X is competitive but still lags behind Nvidia in terms of market share.
|
Feature |
AMD Instinct MI300X |
Nvidia H200 Tensor Core GPU |
|---|---|---|
|
Memory Capacity |
192GB HBM3 |
141GB HBM3e |
|
Memory Bandwidth |
5.3TB/s |
4.8TB/s |
|
Architecture |
AMD CDNA 3 |
Nvidia Hopper |
|
AI Performance (FP8) |
Up to 5,229 TFLOPS with sparsity |
Up to 3,958 TFLOPS with sparsity |
|
HPC Performance (FP64) |
Up to 81.72 TFLOPS |
Up to 34 TFLOPS |
|
Power Consumption (TDP) |
Up to 750W |
Up to 700W |
|
Target Use Case |
Generative AI, HPC, AI training/inference |
Generative AI, HPC, AI training/inference |
AMD Instinct MI300X: The price of the AMD MI300X AI accelerator ranges between $10,000 and $15,000 depending on the customer and volume of purchase. For instance, Microsoft reportedly pays around $10,000 per unit.
Advantages: Higher memory capacity (192GB vs. Nvidia’s 141GB). Solid for large-scale AI models like GPT-3 or PaLM.
Greater memory bandwidth (5.3TB/s) allows faster data transfer for memory-intensive tasks.
Superior peak theoretical performance in AI workloads, particularly in FP8 precision, making it highly efficient for large-scale generative AI tasks.
Higher HPC performance in FP64 precision makes it more suitable for scientific simulations and other high-performance computing tasks.
Disadvantages: Higher power consumption (750W vs. Nvidia’s 700W). But again, this can be offset by the lower cost of the chip itself.
AMD’s ROCm software stack is still behind Nvidia’s CUDA in terms of maturity and optimization for AI workloads. This makes it harder for developers to switch from Nvidia’s established ecosystem.
It also struggles with lower precision tasks like FP8. These are critical for AI training and inference. Nvidia’s H100 and H200 are better optimized here.
Nvidia H200 Tensor Core GPU: The Nvidia H200 is priced significantly higher than the MI300X. It is estimated to cost around $30,000 to $40,000.
Advantages: As I said, it has lower power consumption (700W).
Strong inference performance, particularly in handling large language models like Llama2 with high throughput and efficiency.
Better optimization for specific use cases like long input sequences and batch processing. Chatbots run a lot faster.
Disadvantages: Lower memory capacity and bandwidth compared to the MI300X, which may limit its performance on extremely large models or memory-intensive tasks.
Slightly lower peak theoretical performance in AI workloads compared to the MI300X.
|
Feature/Aspect |
AMD Upcoming Products |
Nvidia Upcoming Products |
|---|---|---|
|
Key Consumer CPUs |
Ryzen X3D Processors |
GeForce RTX 50-Series (Blackwell) |
|
Expected Release Date |
January CES 2025 |
January-March CES/Q1-2025 |
|
Performance Expectations |
Significant gaming performance boost via X3D cache technology |
A major leap in ray tracing and CUDA core count; GDDR7 memory for higher bandwidth |
|
Key Data Center GPUs |
Instinct MI325X |
Blackwell-based AI GPUs |
|
Memory Bandwidth |
Up to 6TB/s (HBM3e) |
Expected improvements over Hopper architecture |
From what I can see: AMD’s MI325X will bring higher memory capacity and bandwidth, making it a strong contender against Nvidia’s upcoming Blackwell-based GPUs. However, Nvidia’s established ecosystem will probably still give it an edge.
It’ll cost much more, though. Blackwell B100 could cost $30,000 to $35,000, whereas MI325X is estimated to be around $15,000-$20,000. Nvidia’s GB200 Superchip could be even more expensive around $60,000 to $70,000.
AMD is “closing” the gap with its upcoming releases—especially in the data center segment—but Nvidia remains dominant across both consumer and enterprise markets right now. I really don’t see it changing anytime soon.
Why?
Nvidia’s CUDA software ecosystem is a major reason customers stick with Nvidia GPUs. CUDA has become the industry standard for AI development. Switching from CUDA to AMD’s ROCm involves significant effort and retraining of teams, and AMD will have to work overtime to convince customers to do that.
Nvidia is growing faster and is more profitable than AMD. Last quarter (Q2), Nvidia grew revenue by 122.4% to $30 billion with a 55.3% net margin. AMD currently has an 11.31% net margin in Q3, and the gap will likely widen once Nvidia reports Q3 figures next week. Here’s what you should know for now:
Note: Q3 numbers for Nvidia are estimates, as per my research.
Valuation is more subjective these days as AI has thrown a curveball into how Wall Street slaps a premium on certain companies. AMD is—and probably will be—valued less than Nvidia, both in terms of market cap and how the broader market sees it. Thus, AMD stock trading at 28 times forward earnings against Nvidia’s 38 times forward earnings doesn’t necessarily make it cheaper.
In fact, I would argue that AMD fares worse here. A lower P/E ratio does not necessarily mean better value. Nvidia’s higher multiple is justified by its dominance in AI chips and its rapid revenue growth in this segment. Nvidia controls between 70% and 95% of the AI chip market, which is expected to grow to $400 billion annually within five years. In contrast, AMD is still playing catch-up in AI—as we looked at before.
What do analysts say?
Nonetheless, I still don’t think AMD is a “better” choice per se. Nvidia has been surprising analysts again and again in the past few quarters. Unless that changes in the coming quarter(s), I don’t think AMD stock offers more value compared to NVDA stock.
In terms of pure value—considering upside potential relative to the current price—AMD stock appears to offer more value than NVDA stock at this moment. NVDA stock remains an excellent long-term play for those focused on stability and leadership in AI, but—if you take price targets seriously—AMD stock provides a better risk-reward balance for investors willing to bet on its continued progress in catching up with Nvidia.
In my opinion, you should hold both if you want exposure to AI, but NVDA stock should comprise more of that pie.
While I acknowledge the potential of both AMD and NVDA as AI plays, my conviction lies in the belief that some AI stocks hold greater promise for delivering higher returns and doing so within a shorter time frame. If you are looking for an AI stock that is more promising than NVDA but that trades at less than 5 times its earnings, check out our report about the cheapest AI stock.
READ NEXT: Why is QUBT Stock up 400%? Is Quantum Computing Inc. Still Worth Buying? and Intel’s Comeback Conundrum: Will INTC Stock Soar or Stumble in 2024?
Disclosure: None.
Add Comment