For longtime shareholders of artificial intelligence (AI) semiconductor designer Marvell Technology (MRVL 3.17%) (myself included), it's been a tough run. The stock has trailed the total return of the chip industry average, as measured by both the iShares Semiconductor ETF and VanEck Semiconductor ETF on a three-, five-, and 10-year basis.
Data by YCharts.
Granted, Marvell has been rebuilding itself under CEO Matthew Murphy, who came over in 2016 from Maxim Integrated (which has since been acquired by Analog Devices). The company made several sizable acquisitions from 2020 to 2022 to get itself positioned for the data center AI boom that is now taking place, and those moves appear to be just beginning to pay off.
Nevertheless, the last earnings report was a mixed bag of results. Is AI going to be a game changer for Marvell, or is it time to cut ties and sell this stock?
Marvell's growing reliance on AI chips
Historically, Marvell was a diversified seller of high-performance network and data storage management chips, including for consumer devices like PCs. Under Murphy, though, Marvell has made a big pivot to data centers, and especially chips used in AI training. The company has drummed up some investor attention by touting its accelerated computing system design partnership with Nvidia, as well as its position as a top competitor to semiconductor giant Broadcom.
In its fiscal 2024 fourth quarter, which ended Feb. 3, the data center business accounted for more than half of total revenue, and on the earnings call, the CEO said that AI chip sales housed within the data center segment landed at "well over $200 million."
The good news is that Marvell expects its data center and AI business to keep growing at a strong pace in the current fiscal year. The bad news is that the rest of its business -- the half not tied to data centers -- is still on the struggle bus.
|
Marvell Segment |
Q4 Fiscal 2024 Revenue |
Change (YOY) |
Q1 Fiscal 2025 Expectations (QOQ) |
|---|---|---|---|
|
Data center |
$765 million |
54% |
Low single-digit percentage increase |
|
Enterprise (non-data center and non-cloud) |
$243 million |
(34%) |
40% decrease |
|
Mobile carriers (5G networks and other) |
$171 million |
(38%) |
50% decrease |
|
Consumer markets |
$143 million |
(21%) |
70% decrease |
|
Automotive and industrial |
$82 million |
(17%) |
Flat |
|
Total revenue |
$1.43 billion |
1% |
$1.15 billion, 20% decrease (down 13% YOY) |
Data source: Marvell Technology Group. YOY = year over year. QOQ = quarter over quarter.
Murphy and the executive team indicated on the Q4 conference call that its non-data center sales could remain weak through the first half of calendar year 2024 (fiscal 2025 for Marvell), offsetting some of its progress in the hot AI market.
Is this an AI hype story to ignore?
The upshot is that, after nearly two years of Marvell working through a period of transition and a semiconductor industry downturn in the wake of the pandemic, the company expects its end markets outside of data centers to finally stabilize and return to sequential growth in the second half of 2024.
Based on the visibility of sales through the next couple of years, Marvell's board of directors authorized a $3 billion stock repurchase plan. Profitability (as measured by free cash flow) has been meager the last couple of years, but this could be the clearest indication that better times lie just ahead.
Data by YCharts.
In fact, Marvell reported a 58% year-over-year jump in free cash flow in the last quarter to $465 million, even as its revenue performance was lackluster. As sales of new AI chips ramp up after several years of development and acquired tech know-how, Marvell could indeed be on the cusp of turning over a new leaf.
Granted, Marvell -- now trading at 56 times trailing-12-month free cash flow -- isn't a cheap stock. But if the company is in fact about to finally return to overall growth this year, selling now could be premature. I'm happy to hang on to my position and see how this AI development cycle pans out.
