When it comes to artificial intelligence (AI) infrastructure, there has been no company as dominant as Nvidia. Its graphics processing units (GPUs) have become the primary chips used to power AI workloads, and the ecosystem the company has built around its chips has created a nice moat.

This starts with its CUDA software platform, which is where most foundational AI code has been written, and then extends to its networking portfolio, where its proprietary NVLink interconnect system essentially helps its chips act as one powerful unit.

Nvidia remains in a strong position, as the company has grown to be the largest in the world, and its revenue ballooned to $57 billion last quarter. At the same time, it is now beginning to see more competition in the space as customers clamor for cheaper alternatives, especially for inference, where Nvidia’s moat isn’t quite as wide, and cost-per-inference is a more important factor since it is an ongoing expense.

Against that backdrop, let’s look at two other AI infrastructure stocks that could have higher upside.

Image source: Getty Images.

Broadcom

One of the biggest sources of increased competition for Nvidia’s GPUs right now is companies beginning to create custom AI ASICs (application-specific integrated circuits). These are preprogrammed chips tailored for specific tasks. While they lack the flexibility of GPUs, they tend to perform well at the jobs for which they were created, and they are generally more energy efficient.

Alphabet‘s tensor processing units (TPUs) have started to gain momentum and are now being offered to its cloud computing customers to run their AI workloads. While Alphabet was responsible for the chip architecture of its TPUs, Broadcom played a big part in turning its designs into physical chips and providing manufacturing support.

Anthropic recently placed a huge $21 billion order with Broadcom for Alphabet’s TPUs to be deployed with Google Cloud. Meanwhile, other companies, including OpenAI, have been turning to Broadcom for its help with their own custom ASICs.

This is a huge opportunity for the company. Citigroup analysts project that Broadcom’s AI revenue will climb from just over $20 billion this past fiscal year to more than $50 billion this current fiscal year and then $100 billion in fiscal 2027.

These estimates also don’t include Apple, with which Broadcom has reportedly been working on its AI ASIC design. With the chipmaker reporting total revenue of just under $64 billion in fiscal 2025, it has an enormous opportunity, making the stock a strong buy.

Taiwan Semiconductor Manufacturing

One of the big advantages of companies working with Broadcom on their custom ASICs is that it has a long-standing relationship with Taiwan Semiconductor Manufacturing, which can provide access to advanced packaging techniques and capacity.

This is important because capacity for advanced chips is currently tight at semiconductor fabs, and TSMC (as the company is also known) is the only foundry that has proved it can manufacture advanced chips at scale with few defects. This has made it one of the most important players in the semiconductor space and is a reason it is a top AI stock.

The great thing for TSMC is that it doesn’t matter whether GPUs or ASICs come to dominate the AI data center landscape; it manufactures both. Each of these types of chips will likely play important roles in AI infrastructure, which will nicely benefit the foundry leader.

TSMC has projected that demand for AI chips will have a mid-40% compound annual growth rate over the next few years, and the company is working to continue increasing capacity to meet this demand. Its position as a virtual monopoly in making advanced chips has also given it strong pricing power, and it is set to continue raising prices for its services in the coming years.

Taken altogether, TSMC is one of the best-positioned companies in the semiconductor space, making it a strong buy right now.