3 Top AI Stocks to Buy Now With Data Center Infrastructure Spending Set to Surge

With Alphabet recently upping its capital expenditure (capex) budget to a whopping $195 billion to $205 billion and saying it plans to spend significantly more on artificial intelligence (AI) infrastructure next year, it has likely opened the floodgates for spending to surge across the industry. While there has been concern about a spending slowdown, there are no signs of one in the near future.

Let's look at three top AI stocks to buy right now on this infrastructure spending surge.

Image source: The Motley Fool.

Nvidia

A combination of a hypergrowth stock trading at a cheap valuation, Nvidia (NASDAQ: NVDA) tops the list of AI infrastructure stocks to buy. The company is the clear leader when it comes to AI model training and has established a wide moat through its CUDA software platform. Most early AI code was written using CUDA and optimized for its powerful graphics processing units (GPUs), which is why it dominates this lucrative market.

However, the company has not sat still, and it also remains well positioned for the inference and agentic AI markets. Its "acquisition" of Groq brought with it language processing units (LPUs) that help enhance its inference offering, while it's also developed its own ARM-based central processing units (CPUs), which play a big role in managing AI agents. At the same time, its networking portfolio has been one of the fastest-growing parts of its business.

Nvidia has continued to see breakneck growth, and the company's transformation into a complete AI infrastructure player that can provide customers end-to-end server solutions for specific AI tasks could be its next big opportunity. With the stock trading at a forward price-to-earnings (P/E) ratio of just 16 times fiscal 2028 (ending January 2028) analyst estimates, this is a great stock to buy while it's cheap, with AI infrastructure spending set to surge.

SK Hynix

With surging AI infrastructure spending, there will be a huge demand for high-bandwidth memory (HBM), which in turn will benefit SK Hynix (NASDAQ: SKHY). The Korean company invented HBM and is the leader in the space with over 50% market share. It is also the main supplier for Nvidia and the second-largest supplier for Alphabet's tensor processing units (TPUs).

HBM is a specialized form of DRAM (dynamic random-access memory) that gets packaged with GPUs and other AI chips. Right now, demand for HBM is surging while supply remains tight, boosting not just HBM prices but also DRAM pricing. The reason is that the big three DRAM makers are focused on HBM, and they can only increase capacity so much. HBM and AI chips are both manufactured using EUV (extreme ultraviolet lithography) machines, and only one company in the world, ASML, makes them. At the same time, HBM uses around three times the wafer capacity of ordinary DRAM, further constraining supply.

SK Hynix has said the DRAM industry will see its largest-ever supply-demand imbalance next year and that the market will likely remain supply-constrained until at least 2030. This dynamic has allowed the company to sign long-term supply contracts with no price caps, positioning it to benefit from further DRAM price increases while also increasing its visibility. With the stock trading at a forward P/E of around 8, it looks like a top buy on surging AI infrastructure demand.

Broadcom

As the co-developer of Alphabet's TPUs, Broadcom (NASDAQ: AVGO) will directly benefit from the search giant's increased AI data center spending. Meanwhile, other hyperscalers (owners of large data centers) have also turned to it to make their own custom chips, including Meta Platforms and OpenAI, while Anthropic has a deal in place to buy TPUs. The company has projected that its custom AI chip business will exceed $100 billion in fiscal 2027, and that number may be low.

Meanwhile, Broadcom is also a leader in data center networking, a fast-growing business closely tied to AI infrastructure spending. With its non-AI business also poised to rebound, the company is poised to see massive growth in the coming years.

Trading at an attractive forward P/E of 19.5 times the fiscal 2027 consensus, this is a top AI infrastructure stock to buy.

Geoffrey Seiler has positions in Alphabet, Broadcom, and Meta Platforms. The Motley Fool has positions in and recommends ASML, Alphabet, Arm Holdings, Broadcom, Meta Platforms, and Nvidia. The Motley Fool has a disclosure policy.