焦
The artificial intelligence (AI) build-out has made chip companies some of the fastest-growing large businesses in the world. Nvidia ( NVDA +1.34% ) grew revenue 106% year over year in its most recent quarter. Broadcom ( AVGO +2.97% ) grew 86%.
Micron Technology ( MU +3.92% ) more than quadrupled its sales. Even Marvell Technology ( MRVL +1.45% ) , the group's smallest, grew 37%. But holding a stock through 2030 is a different bet than chasing last quarter's growth.
AI spending will likely cool at some point over the next four years. What matters then is where in the chip supply chain the profits hold up. Here's what each one has to keep doing for a four-year hold to work.
Image source: Getty Images. Nvidia: the platform Nvidia's revenue in the fiscal second quarter of 2027 (ended July 26) rose 106% year over year to $96.2 billion, with $89 billion of that coming from data centers. In late August, management told analysts it expects revenue to grow approximately 70% in fiscal 2028 -- an outlook it called limited by supply, not demand.
At about $212 as of this writing, shares trade at about 14 times fiscal 2028's estimated earnings. That price, I'd argue, stops giving Nvidia credit after next year. The job through 2030 is to stay the platform AI computing runs on (the chips, the networking, and the software) even as its biggest customers fund custom alternatives.
Premium Feature Moneyball Superscore 94 /100 Today's Change ( 1.34 %) $ 2.93 Current Price $ 222.27 Broadcom: the custom-chip franchise Broadcom is the company building many of those alternatives. Its AI semiconductor revenue reached $16.7 billion in the fiscal third quarter of 2026 (ended Aug. 2), 221% higher than a year earlier -- and still accelerating, with growth of 54% from the prior quarter alone. Custom accelerators (chips designed with a single customer) made up 73% of it.
CEO Hock Tan told analysts earlier this month that Broadcom has secured the supply to double AI revenue to about $115 billion in fiscal 2027 -- and to $230 billion the year after. The stock trades at about 18 times fiscal 2027's expected earnings. The four-year job here is delivery: a short list of AI labs and cloud giants taking billions of dollars of chips on schedule, year after year.
Premium Feature Moneyball Superscore 90 /100 Today's Change ( 2.97 %) $ 10.31 Current Price $ 357.61 Marvell: the aggressive bet Marvell runs the same strategy at a fraction of Broadcom's scale. Revenue reached $2.7 billion in its fiscal second quarter of 2027 (ended Aug. 1), and data center revenue grew 46%. And CEO Matt Murphy now expects about $18 billion of revenue in fiscal 2028, up from around $12 billion this fiscal year, with the custom business more than doubling along the way.
Marvell is also the most expensive of the four, at about 33 times its expected fiscal 2028 earnings. That price-to-earnings multiple only works if the design wins behind those forecasts keep ramping on time. Premium Feature Moneyball Superscore 88 /100 Today's Change ( 1.45 %) $ 3.49 Current Price $ 244.25 Micron: the cheapest, and the most cyclical Micron sells the memory all of that AI hardware depends on, and right now that is arguably the supply chain's most profitable place.
Revenue in its fiscal third quarter of 2026 (ended May 28) came to $41.5 billion, over four times the year-earlier total. Its non-GAAP (adjusted) gross margin reached 84.9%, up from 39% a year earlier. And guidance for the fiscal fourth quarter, which ended in early September, calls for about $50 billion of revenue.
Yet at about $928 as of this writing, shares cost only about 6 times analysts' fiscal 2027 earnings estimates, the group's lowest price-to-earnings multiple by far. In other words, investors are already betting these margins shrink. After all, memory prices have cycled for decades, and I think some of that doubt is fair.
Selling a business earning at this level seems premature. But this deep into memory's profitable stretch, I wouldn't put new money in at today's price, either. The ones I'd buy today Which of these, then, would I buy and hold through 2030?
Nvidia and Broadcom are my anchors. Both sit where the supply chain's most durable profits appear to be. And both can grow into their prices if most of the demand they have guided for shows up.
Marvell is a buy as well, though I'd keep it the smallest of the three, because more of its growth still has to convert from bookings into revenue. Of course, if AI spending turns down hard before 2030, all four stocks could fall together, whatever their valuation multiples. All four fell between 5% and 10% in a single week this month, so a four-year holder should expect stretches far worse than that.
I'd buy Nvidia, Broadcom, and Marvell today. Micron, I view as a hold.
Source: The Motley Fool
