AMD tops $1 trillion and joins the club once reserved for Nvidia

AMD topped the $1 trillion mark in market value for the first time this Monday, the 21st. The figure is equivalent to about R$5.14 trillion. Shares rose 9.6% during the morning, reaching $613.50, the highest price in the company's history.

The driver behind the rally has a familiar name. Investors are betting on the company's growth in AI-focused computing.

With that, the company became the fourth American chipmaker to reach that level. Before it, Nvidia, Broadcom, and Micron had already gotten there.

AMD up 26% in less than a month

The rally started in early September. At the time, the company released a quarterly revenue forecast above Wall Street estimates.

Curiously, the figure fell short of investors' most optimistic expectations. Even so, shares have gained more than 26% since then.

In addition, the move tracked the entire sector. This Monday, Intel rose about 11% and Qualcomm gained 4.1%. Likewise, the U.S. semiconductor index climbed 2.6%, its highest level in a month.

The strategy of selling the whole system

The company is seen as Nvidia's main rival in AI GPUs. However, the game has changed in recent years.

Today, AMD sells much more than standalone chips. It offers complete systems, combining processors, networking equipment, and other components.

That choice has a clear logic. After all, whoever builds an AI cluster buys integrated infrastructure, and Nvidia had already been selling that package for some time.

The company also gains from CPUs. AI servers combine central processors with GPUs, and that demand has helped the company take share from Intel.

AMD still chasing a $5 trillion rival

The size of the gap is striking. Nvidia reached the trillion mark in 2023 and now stands above $5 trillion.

Therefore, AMD has joined the club, but it remains well behind the leader. The value gap largely reflects the competitor's ecosystem advantage.

This point directly matters to those who write code. CUDA has become the de facto standard for training and inference. AMD, on the other hand, is betting on ROCm, its open software stack for GPUs.

What this shift changes for developers

First, real competition tends to lower the price of computing. More GPU supply tends to ease the cost per hour in the cloud.

Second, portability gains value. Code tied to a single stack limits hardware choice in the future.

For that reason, it's worth favoring high-level frameworks. PyTorch, JAX, and inference runtimes already abstract away much of the differences between vendors.

Also, test your models on more than one backend. Differences in performance and numerical precision show up early in this kind of validation.

Also consider the complete system. Network, memory, and interconnect matter as much as the GPU in distributed workloads.

Finally, keep an eye on library maturity. Optimized kernels and quantization support determine whether switching hardware pays off.

Supercomputers and the deal with the U.S. government

In October of last year, the U.S. government announced a $1 billion partnership with AMD. The goal is to develop supercomputers for scientific research.

The areas mentioned include nuclear energy, cancer treatments, and national security. According to Energy Secretary Chris Wright, the AI in these systems could accelerate advances in nuclear fusion within the next two or three years.

What to watch in the coming months

Keep an eye on the company's next quarterly earnings report. It will show whether data center revenue keeps pace with the market's euphoria.

Also follow the evolution of ROCm and the availability of AMD GPUs on the major clouds. In the meantime, it's worth keeping your code ready to run on more than one vendor.

Follow our profile on Instagram!