What the Rise of Generative AI for Networking Means for Advisors

The world of artificial intelligence (AI) is rapidly evolving. With surging hype around AI, our optimism around generative AI networking is long term—in fact, generative AI networking equipment spending is expected to reach $34 billion in 2028, up from our estimate of $8 billion market in 2023.

Using generative AI can also help financial advisors improve client communication and build portfolios that capitalize on AI investment opportunities. Morningstar researchers examine AI networking, market forecasts, analyst top picks, and more.

To read the full research report, download a copy.

One of the powers of generative AI is its ability to create content in different tones, reading levels, and contexts. However, training a new language model using generative AI is expensive—we believe a model like GPT-4 took roughly three months and $100 million to train, yet estimate $30 million of that was wasted due to graphics processing units (GPUs) sitting idle waiting for data.

Ultimately, the progression of generative AI models will require faster, more efficient networks. The end goal for these model builders is speed: the speed to train a model and the speed to give outputs for users during inference. But networking speed advancement is difficult, which is why specialized networking chip and equipment designers are important for generative AI model builders—and why we rate the top tier of these providers as moaty, with hardware and software design expertise that’s hard to replicate.

Here's what advisors need to know about generative AI networking.

Morningstar researchers forecast 34% compound annual growth for the generative AI networking market through 2028. This also matches our views for generative AI accelerator chip growth. Our estimate assumes spending toward generative AI networking will quickly become a large majority of overall high-speed data center networking spending—making up roughly 75% of total high-speed data center networking spending in 2028.

Some companies will benefit disproportionately from this surge in spending. These top players have carved economic moats with prowess in the design of cutting-edge networking chips and equipment and used these moats to establish strong market shares.

Our 2023 market estimate implies that generative AI already makes up nearly half of overall high-speed data center networking spending. We believe generative AI networks are almost exclusively using speeds of 400 Gbps—which is the highest level in volume production today.

This high weighting toward generative AI demonstrates three key tenets:

1. Generative AI spending has already grown to a significant figure in the past two years.
2. Generative AI networking requires the most cutting-edge hardware—which is the most expensive—and a lot of it.
3. A substitution effect is occurring within high-speed data center networking.

More specifically, cloud providers and large data center builders are prioritizing building out generative AI capacity over building out non-AI infrastructure. While we continue to expect growth for non-AI high-speed networking, we think generative AI will drive overall high-speed networking growth over the next five years.

Arista, Broadcom, Marvell, and Nvidia will be the primary beneficiaries of spending on generative AI networking. Investors should look to these stocks as ways to invest in rising generative AI networking spending—Morningstar rates all four as moaty and high quality.

When it comes to these companies, we also foresee generative AI accounting for a significant portion of sales by 2028. For Nvidia and Broadcom, the AI networking impact is meaningful, but smaller than the AI accelerator opportunity (and in the case of Broadcom, its software exposure). Marvell has a well-balanced opportunity between its networking and accelerator sales for generative AI. We view Arista as the purest play on generative AI networking, with sales from generative AI networking specifically growing to more than half of the top line in five years within our forecast.

Marvell is a smaller player in networking chips, but we consider it to be well positioned to rapidly gain share with a portfolio tailored to high-speed data center networking and AI. We believe investors looking for a generative AI networking chip play have favored Broadcom over the past year, leading to an opportunity for Marvell.

The company has two design wins with Amazon Web Services for its Trainium and Inferentia chips. And it recently acquired a second custom AI chip customer, which we suspect is Microsoft. These wins give us confidence in our aggressive forecast for AI custom accelerator revenue to jump to $2.5 billion in fiscal 2029, up from nearly nothing in fiscal 2024. Within optical chips, Marvell also serves pretty much every major data center in the world with its commanding market share.

Over the next several years, AI is expected to be the key growth driver across the broad market. In fact, our recent report on semiconductor industry trends reveals AI accelerator (for example, chip) revenue will increase roughly four times in this time frame.

And industry leaders like Nvidia won’t be the only beneficiary. Specifically, we expect AI to be nearly 40%-50% of total company revenue for firms including Broadcom, AMD, and Marvell Technology.

There are other parts of the value chain that will benefit beyond processing chips—we see material growth coming for networking firms (chips and ethernet connections), memory firms (high-bandwidth memory), and equipment firms and foundries (rising complexity).

Yet when it comes to the rapid AI growth in semiconductors, Nvidia still plays a major role. Our latest semiconductor industry market analysis finds that the historic rise in demand for AI and accelerators is particularly led by Nvidia's data center GPUs, which will most likely continue in 2025. We also anticipate an ongoing shift in spending toward AI-empowered data center gear and away from traditional equipment.

That’s not all: On an annual basis, Nvidia’s GPU revenue should rise more than 30 times in just six years. Plus, the company extracts tremendous pricing power on its data center products—reaching gross margins as high as 78%.

Concerns of tariffs, export restrictions, and prolonged weakness in the automotive and industrial markets have tempered valuations in the Asia semiconductor landscape. Our Q4 2024 manufacturing report shows that Chinese semiconductor stocks are now expensive—with Taiwan Semiconductor Manufacturing Company (TSMC), Samsung, and most wafer producers undervalued in our view.

Specifically, TSMC’s advantage stems from a cycle of technology lead, research, and premium pricing. As for Samsung, we believe it should eventually be qualified by Nvidia to supply high-bandwidth memory (HBM). GlobalWafers’ efficient operations—with sites in Asia, Europe, and the US—are also key to its superior gross margin.

When it comes to the cybersecurity landscape, AI will be a new demand driver as attacks become more sophisticated. While the AI security market is small right now, we expect rapid growth as the use of applications proliferate. And for good reason—our recent cybersecurity report suggests the cost of a data breach is almost 50% higher for companies not using AI security compared to those that leverage these tools. Given that roughly 10%-12% of public cloud spending is geared toward security solutions, we forecast that the public cloud spending catalyzed by AI will add another $15-$18 billion in security spending.

Generative AI can be a useful tool for advisors to strengthen client relationships and show value. When you have a deeper understanding of the landscape, it can be easier to identify the right AI stocks for clients and meet investors’ evolving needs.

Streamline your work process with Mo, our AI chatbot powered by Morningstar Intelligence Engine. Drawing from our extensive investment data and research library—including every Morningstar.com article—Mo is designed to quickly surface and summarize our independent insights in a conversational format for clients.