There has been a lot of talk about artificial intelligence and the impact it will have on humanity and the economy ever since the release of ChatGPT last November. But that enthusiasm began to translate into eye-opening financial results with Nvidia's (NVDA 4.35%) blowout guidance on May 24 -- roughly six months later.
Nvidia's guidance seemed to validate the recent bullishness for semiconductors and cloud platforms generally, propelling many such stocks significantly higher. However, for some, Nvidia may now look at tad pricey, at over 200 times earnings and 35 times sales.
Image source: Getty Images.
Fortunately, there are other data center-oriented chip stocks that will benefit handsomely from the AI trend, and they come at a cheaper valuation -- like the following three stocks.
Broadcom and Marvell
I group these two companies together because they have some overlap in terms of the products they offer, including Ethernet switching, optical component interfaces, storage controllers, and customized application-specific integrated chips (ASICs).
All of these chips will be essential in building out next-generation data centers geared for AI applications. On the company's recent conference call with analysts, Marvell (MRVL 1.21%) CEO Matt Murphy said:
... [L]eading cloud data centers connect thousands of these systems in a single cluster to provide maximum scalability for their customers, with each of these systems capable of driving tens of terabits of network traffic... And in order to create the largest possible cluster sizes at data center scale, these connections need to be able to operate over increasingly long distances. These clusters require a staggering amount of high-bandwidth connectivity, all of which needs to be provided at ultra-low latency and high reliability and within a reasonable power envelope.
In other words, along with GPU-based computing systems, networking and switching requirements are also about to skyrocket, which will play into the hands of Marvell's PAM4 digital signal processing (DSP) interfaces, data center interconnect, and Teralynx 10 Ethernet switching platforms.
Meanwhile, Broadcom's (AVGO 1.64%) current open architecture Tomahawk and Jericho chips for merchant switching and routing, currently deployed in data centers and telecom platforms, can be easily adapted and used for the intense networking needs of AI data centers.
In addition to these routing, switching, and optical interconnect platforms, both companies also compete in the arena of custom ASICs, which many large cloud companies are using with their custom-designed AI accelerators. JPMorgan Chase analyst Harlan Sur believes this will be a fast-growing market, with Broadcom having a leading 35% share, and Marvell having roughly 15% ASIC share. Yet, Sur sees both as poised to benefit from custom AI design wins in the years ahead.
Overall, both companies recently stated they see their AI-related revenue doubling this year, with another doubling next year. Broadcom notes AI-related products represent about 15% of its semiconductor business today, but remember, Broadcom also has a software segment that accounts for about 22% of revenues, so AI chips account for a little less than that 15% overall.
Meanwhile, for Marvell, AI-related products only accounted for $200 million of revenue last year out of $5.9 billion in revenue, or about 3.3%; however, Marvell obviously sees very strong growth here. Moreover, management isn't accounting for the knock-on benefits for its large storage control portfolio in its AI growth projections, which should benefit from increased data usage. Thus, the doubling in 2023 and 2024 AI revenue could understate the effect of AI on Marvell's business.
Broadcom is the more mature and profitable company, trading at just 18 times forward earnings with a dividend of 2.3%. Meanwhile, Marvell is smaller, at just a $51.8 billion market cap to Broadcom's $338 billion. Marvell isn't currently profitable on a generally accepted accounting principles (GAAP) basis, but is free cash flow profitable, due to high stock-based compensation.
The stock trades at 40 times this year's earnings expectations, and about 25 times 2025 analyst estimates, as analysts expect profitability to inflect upwards.
And don't sleep on Microchip
Meanwhile, Microchip Technology (MCHP 2.27%) also isn't the first chip stock many think about when they mention AI. This could be due to the company's microcontroller and analog focus, which are very diverse businesses beyond data center infrastructure.
However, Microchip's solutions are important to the data center, and management has identified the data center as one of six big megatrends it cited and attacked as part of its Microchip 3.0 strategy launched two years ago. In fact, Microchip's data center segment is the largest of the six megatrends cited, at 17.5% of sales, up from 14.2% just two years ago. Overall, Microchip's high-growth megatrend segments have grown from 34% to 45% of the business in just two years, growing at two times the rate of the overall company.
While not every Microchip data center product plays a role in AI, many do, including its PCIe switches, which connect servers and individual chips to each other as opposed to whole-rack systems like Ethernet does. Other AI-related products include smart memory interfaces, data center interconnect Ethernet products, power and silicon carbide chips, and others. All should benefit from increasing compute intensity, and Microchip's focus on sustainability and electrification should help mitigate the heat generated by massive AI clusters going forward, too.
Microchip is the cheapest of the three stocks, trading at just 11.5 this year's earnings estimates. While its AI impact may be more muted than the other two, Microchip's low valuation and increasing cash returns give it a lower hurdle to clear. Meanwhile, the rest of its portfolio is geared heavily toward industrial, IoT, and automotive chips, which should also have resilient growth as the electrification and automation trends sustain through this decade.
