With spending for artificial intelligence (AI) infrastructure still on the rise, now looks like a good time to invest in the space. Let's look at three stocks set to be AI infrastructure winners.
Nvidia
Nvidia (NVDA 4.31%) remains the king of AI infrastructure, and it proved that once again when it recently reported a 63% surge in its fiscal third-quarter revenue to $57 billion. That type of growth at a company the size of Nvidia is just unheard of.
NASDAQ: NVDA
Key Data Points
Demand for its graphics processing units (GPUs) remains insatiable, but that's not even the fastest-growing part of the company. Nvidia's data center networking portfolio -- which consists of its NVLink interconnect system, InfiniBand, and Spectrum-X Ethernet products -- grew its revenue by a whopping 162% to $8.2 billion last quarter, as the company is now selling complete end-to-end AI solutions, which it calls AI factories.
Nvidia continues to have a wide moat in the AI accelerator space with its CUDA software platform, which is where most foundational AI code is written. Meanwhile, its NVLink interconnect solution helps its GPUs act as one powerful unit, which keeps customers from mixing and matching chips within an AI cluster.
If AI spending continues at its rapid pace, which appears likely, Nvidia is a stock to own.
Image source: Getty Images.
Broadcom and AMD
While Nvidia dominates the AI chip market, there is an opening for both Broadcom (AVGO 0.60%) and Advanced Micro Devices (AMD 7.36%) to gain some traction, especially as inference becomes more important. Unlike training, inference is an ongoing cost, so cost per inference and total cost of ownership become much more important factors. Inference also isn't quite as technically challenging as training large language models (LLMs), so Nvidia's CUDA moat isn't quite as wide.
Broadcom and AMD are going after the AI infrastructure market in two very different ways. Broadcom has a strong networking portfolio of components that are essential in helping manage the flow of data and distributing AI workloads across servers. This is a strongly growing business, but its biggest opportunity is in helping hyperscalers (companies that own large data centers) design their own custom AI chips called ASICs (application-specific integrated circuits).
NASDAQ: AVGO
Key Data Points
ASICs are pre-programmed chips that are designed for specific tasks. One of the big advantages of ASICs is that they tend to consume less power than GPUs, which can make them more cost-effective, especially when it comes to inference. However, the big disadvantage is that since they are pre-programmed, they lack the flexibility of GPUs and can't be optimized if technology shifts.
Broadcom helped Alphabet design its highly regarded Tensor Processing Units (TPUs), which has led to other hyperscalers flocking to it to help them design their own AI ASICs. Broadcom has said its three customers further along represent a $60 billion to $90 billion market opportunity in its fiscal 2027, while a fourth customer just placed a $10 billion order for next year. In addition, it just signed a large deal with OpenAI.
These deals should help power strong growth for the company over the next several years, making it a solid stock to own.
NASDAQ: AMD
Key Data Points
AMD, meanwhile, is the market leader in data center central processing units (CPUs). While GPUs provide the power, CPUs provide the brain power. The market is growing quickly with the data center build-out, and AMD has been taking share.
However, AMD's biggest opportunity is grabbing a bigger slice of the GPU market as inference becomes more important. It has established a niche in this market, and there are developments that indicate it can take some share. It has also entered into a partnership with OpenAI to provide compute power, but perhaps an even bigger deal is that Microsoft is developing a tool kit to convert Nvidia CUDA code to AMD's ROCm software platform for inference.
AMD recently set out ambitious long-term targets, and stock has solid upside even if it can just gain a small amount of share in the big and growing AI accelerator market.
