Spending on artificial intelligence (AI) infrastructure continues to rise, setting the stage for companies in the AI chip space to outperform over the coming years. The growth in AI infrastructure is being led by a combination of cloud computing operators investing heavily to keep up with demand, AI start-ups and large tech companies racing to build out the best foundational large language models (LLMs), and even countries investing in AI.
Five of the best AI semiconductor companies to invest in right now to profit from this trend are Nvidia (NVDA +1.78%), Advanced Micro Devices (AMD +1.19%), Broadcom (AVGO +0.59%), Marvell Technology (MRVL +1.89%), and Taiwan Semiconductor Manufacturing (TSM +0.63%).
Let's look at what each brings to the table.
Image source: Getty Images.
Nvidia
Nvidia's graphics processing units (GPUs) have become the backbone of AI infrastructure, providing the power needed to run AI workloads. The strength of GPUs comes from their ability to use parallel processing, which simply means they can perform many tasks at once. This is ideal for AI because the technology requires tons of calculations to be done quickly.
NASDAQ: NVDA
Key Data Points
Nvidia's wide moat, however, comes from its CUDA software platform. GPUs were originally designed to speed up graphics rendering in video games, and Nvidia created CUDA as a way to allow developers to program its chips for other tasks. With the market outside of gaming slow to develop, rivals were slow to develop their own software platforms. Nvidia pushed CUDA into universities and research labs, where developers were taught on its systems. It has since built out tools and libraries on top of CUDA to improve the performance of its GPUs, giving it a big advantage in the space. In the first quarter, it had a whopping 92% market share in the GPU space.
As such, where AI infrastructure spending heads, Nvidia is sure to follow.
Advanced Micro Devices
As a distant No. 2 player to Nvidia in the GPU market, AMD has established itself as a leader in central processing units (CPUs) used in the data center space. While GPUs provide the power, CPUs act as the brains. It's a solid, growing market, but much smaller than the one for GPUs.
NASDAQ: AMD
Key Data Points
However, the company has carved out a solid niche with its GPUs in the AI inference market, where performance demands are lower and cost becomes a bigger factor. That helps level the playing field with Nvidia, whose CUDA advantage is less critical here. With inference expected to become the (much) larger of the two markets over time, AMD has a lot of growth potential, even if it can only take a little market share away from Nvidia in the future.
Broadcom
Broadcom is a networking leader whose technology helps improve the flow of data across data centers and AI clusters. Its Ethernet switches and other components are essential for handling the massive data traffic in high-performance computing environments, so that everything runs smoothly and does not get backed up. As AI clusters get bigger, its networking portfolio becomes even more valuable, given its role in managing congestion and scaling bandwidth efficiently.
That said, while its networking portfolio is currently the main driver of its growth, custom AI chips are its biggest opportunity. Broadcom played a critical role in helping Alphabet develop its custom Tensor Processing Units (TPUs), and that success has led to an increasing list of custom AI chip customers. While most of these chip customers are in the early stages, the company sees its three furthest-along custom AI chip customers as a $60 billion to $90 billion serviceable market opportunity in fiscal 2027. While it won't capture all of this, it's a huge growth opportunity that doesn't even include newer customers, such as Apple.
Marvell Technology
Broadcom isn't the only company helping customers build custom AI chips. Marvell has contributed IP (intellectual property) and interconnect technology used in Amazon's custom chips like Graviton and Trainium. While there's some concern Amazon could eventually look to bring more of the work in-house or shift to other partners, Marvell expects its broader relationship to grow. In addition to providing IP for its custom chips, Marvell also reportedly supplies networking chips, connectivity solutions, and storage controllers to Amazon, which are key components for scaling AI infrastructure.
NASDAQ: MRVL
Key Data Points
Marvell is also believed to be providing IP to Microsoft's new custom chip, Maia. J.P. Morgan analysts noted that it recently won a deal to be part of future generations of the new chip. This program is still in its very early stages, but it could be a significant revenue growth driver in the future.
Taiwan Semiconductor Manufacturing
While the above companies are involved in the design of chips, Taiwan Semiconductor Manufacturing is the one generally making them. As such, it doesn't matter who is the biggest AI chip winner or who takes market share -- Taiwan Semiconductor wins as long as AI infrastructure and chip spending continue to grow.
Because of its technological expertise and scale, and the struggles of its main competitors, TSMC has become the world's primary manufacturer of advanced chips. It's been seeing strong growth, driven by capacity expansion and price increases. TSMC is working closely with its largest customers to help meet growing chip demand, which should pave the way for years of strong growth ahead.
