3 Stocks to Buy as AI Infrastructure Continues to Surge

With Alphabet and Amazon recently upping their 2026 capital expenditure (capex) budgets and signaling they will spend even more in 2027, now may be a good time to invest in stocks set to benefit from this spending spree. Best of all, many of these stocks are trading well off their highs after earlier fears that artificial intelligence (AI) infrastructure capex might slow, which it now shows no signs of doing.

Let's look at three top AI stocks to buy right now.

1. Nvidia

Nvidia (NASDAQ: NVDA) remains one of the best ways to play the infrastructure boom. The company continues to see rapid growth, and the stock is cheap, trading at a forward price-to-earnings (P/E) ratio of just 16 times fiscal 2028 (ending January 2028) analyst estimates.

Although the AI infrastructure market is evolving, Nvidia has positioned itself to remain a leader. Its graphics processing units (GPUs) remain the dominant systems used in AI model training, while its CUDA software, in which most foundational AI code is written, gives it a wide competitive moat.

Meanwhile, Nvidia's networking portfolio has become another growth driver as chip cluster sizes grow, providing the nervous system for its end-to-end AI server solutions. It's also developed other chips, such as central processing units (CPUs), which are becoming increasingly important for agentic AI, and data processing units (DPUs) that go into these systems.

Perhaps the company's smartest move, though, was its acquisition of Groq and its language processing units (LPUs). LPUs have SRAM (static random-access memory) embedded directly on the chips, making them especially useful for the decode phase of inference.

In contrast, its GPUs, packaged with high-bandwidth memory (HBM), can more cheaply handle the pre-fill phase. It's a nice solution that positions it to be a leader in inference, which is expected to become an even larger market than training.

Image source: Getty Images.

2. SK Hynix

One of the biggest bottlenecks in the AI infrastructure segment remains HBM, as GPUs and other chips need to be packaged with it to reduce latency and improve power efficiency. HBM is even more important with inference, which is further driving demand.

However, the market remains very supply-constrained, as HBM requires three times the wafer space as ordinary DRAM, and the big three DRAM (dynamic random-access memory) makers are competing for the same machines that are also needed to manufacture advanced logic chips, such as GPUs.

As the HBM market leader, SK Hynix (NASDAQ: SKHY) is the memory company best positioned for the long term. It has more than 50% market share and is the primary HBM supplier to Nvidia, with whom it just signed a major $500 billion multiyear supply deal. With the DRAM market expected to be in short supply during the next several years and the company signing long-term contracts with no price caps, it looks like a solid buy, with the stock trading at a forward P/E of a little more than 5.

3. Taiwan Semiconductor Manufacturing

Manufacturing advanced chips is technically challenging, and the only company that has proven it can consistently produce them at high yields (few defects) at scale is Taiwan Semiconductor Manufacturing (NYSE: TSM), also known as TSMC. This has given TSMC a virtual monopoly in the sector and has made it one of the most important players in the semiconductor ecosystem. It has also given the company strong pricing power, helping it achieve robust gross margins.

With the proliferation of chips in AI data centers, TSMC is uniquely positioned to be a big winner. It's working to aggressively build capacity to meet surging demand, and as a key partner to its customers, it has some of the deepest ties within the industry. With the stock trading at 19 times 2027 analyst estimates, it is attractively valued given its growth outlook.

Geoffrey Seiler has positions in Alphabet and Amazon. The Motley Fool has positions in and recommends Alphabet, Amazon, Nvidia, and Taiwan Semiconductor Manufacturing. The Motley Fool has a disclosure policy.