Got $1,000? 2 Growth Stocks Building the Data Centers of the AI Infrastructure Supercycle.

If you have $1,000 that you're ready to invest today, that would be enough to start positions in two of the chip stocks that look poised to lead the next wave of the artificial intelligence (AI) build-out in Advanced Micro Devices (NASDAQ: AMD) and Broadcom (NASDAQ: AVGO). With $1,000, you could buy one share of each.

Let's look at why these two semiconductor stocks look like solid investments.

1. AMD: An inference, agentic AI, and physical AI winner

AMD was outmaneuvered by Nvidia in the AI training market, but it's positioning itself to prevent that from happening again. It's developing processors that target the inference, agentic AI, and physical AI markets, and aims to be a viable contender in all three.

It has also been making significant investments in memory. This positions it well for the inference market, where architectures that prioritize high volumes of fast memory access are favored over raw compute power.

The chiplet design of its graphics processing units (GPUs) allows them to be packaged with more memory, and AMD has also formed a partnership with Cerebras to offer a disaggregated inference solution that combines both of their systems. It has also inked a deal to acquire inference chipmaker Taalas, which hardwires AI models directly into its chips. It also bought memory optimization company MEXT earlier this year. And AMD has two $100 billion inference deals in place with OpenAI and Meta Platforms, as well as a large agreement with Anthropic.

On top of that, the company is a leader in the data center central processing unit (CPU) market. This market has been exploding with the rise of agentic AI, as servers dedicated to powering AI agents require considerably more CPUs than those used for AI training. AMD forecasts that the data center CPU market will grow to $220 billion in the coming years, although in light of the rapidly growing popularity of Meta's new Muse agentic AI offering, this estimate could prove low. Meta, by the way, is running its Muse agents on AMD's CPUs.

Finally, with its agreement to acquire AI model and research lab World Labs, AMD is improving its ability to pursue the physical AI and robotics markets. World Lab's expertise lies in spatial intelligence models, and it will provide AMD with the tools to advance an open AI ecosystem (in contrast to Nvidia's closed system) that serves those emerging AI markets. This deal isn't about selling more chips in the near term, but rather about creating future architectures and platforms for physical AI.

Given all its opportunities for growth ahead, AMD stock is a buy.

Image source: The Motley Fool.

2. Broadcom: The custom chip giant

With AI infrastructure spending soaring, hyperscalers (owners of large data centers) are increasingly looking to reduce their costs, and one way they're doing that is by seeking other sources of processing power than GPU behemoth Nvidia. One way they are doing this is by designing their own custom application-specific integrated circuits (ASICs). These are purpose-built, hardwired AI accelerators that handle the narrow workloads for which they were designed more efficiently than a general-purpose GPU can, and consume less power while doing so.

However, going from design concepts to physical chips that can be manufactured at scale is not a simple process, which is where Broadcom comes in. It is the leader in ASIC technology and provides important intellectual property and services, collaborating on designs with its tech sector clients. It helped Alphabet develop its well-regarded Tensor Processing Units (TPUs), and the success those have found has helped induce other hyperscalers to sign on with Broadcom.

TPUs will remain Broadcom's biggest business over the next few years, as both Alphabet and Anthropic are buying them. In fact, management says the AI lab is set to become its largest XPU customer next year. It has also helped Meta and OpenAI create their own custom chips; production of those is just starting to ramp up. As such, management has forecast that its custom chip business revenues will double next fiscal year to $115 billion, and then double again the following year to $230 billion.

On top of that, Broadcom is also a leader in data center networking. It makes Ethernet switches, digital signal processors (DSPs), SerDes (Serializer/Deserializer), and network interface cards (NICs) that data centers use to distribute AI workloads and transfer data across servers. Its biggest strength lies in AI data center Ethernet solutions, powered by its Tomahawk and Jericho chipsets. As AI server clusters continue to grow in size, this is becoming a huge business, and directly ties into its custom chip unit.

Broadcom is set to see explosive growth in the years ahead, yet the stock trades at a cheap valuation of less than 12 times its estimated fiscal 2028 earnings. This makes the AI stock a great buy.

Geoffrey Seiler has positions in Advanced Micro Devices, Alphabet, Broadcom, and Meta Platforms. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Broadcom, Meta Platforms, and Nvidia. The Motley Fool has a disclosure policy.