Sell these "safe" blue chips immediately (From Chaikin Analytics)

NVIDIA (NASDAQ: NVDA) just cleared the bar, and then some. On Wednesday, Aug. 26, the AI chip leader reported blowout results, with revenue of $96.2 billion and data center sales up 117% year over year. The company also guided current-quarter revenue to a stunning $108 billion, well above Wall Street's expectations. CEO Jensen Huang declared that AI has "reached its inflection point" and that demand is accelerating.

Now the market gets its next read. When Marvell Technology (NASDAQ: MRVL) reports Thursday, Aug. 27, after the close, one day after NVIDIA, it will help answer a more nuanced question: whether the AI infrastructure story is genuinely broadening beyond the GPU, into the custom silicon, networking, and interconnect layers the entire buildout depends on.

Marvell Tests the AI Buildout Beyond GPUs

It is easy to forget that an AI data center is far more than a room full of GPUs. Those chips need to talk to one another at staggering speeds, move data across racks and between facilities, and connect to storage and the outside world. That is Marvell's territory. The company is a leader in optical interconnects, high-speed SerDes, Ethernet switching, and custom silicon that stitches AI clusters together.

Marvell has become one of the more credible AI infrastructure names outside the GPU leaders. Its data center segment has been growing rapidly as hyperscalers pour capital into their networks, and NVIDIA's accelerating demand outlook only strengthens the backdrop. That data center growth is the number investors will scrutinize most closely on Thursday.

The Custom Silicon Opportunity Just Got Bigger

The most compelling part of the Marvell story right now is custom silicon. Hyperscalers increasingly want their own tailor-made chips rather than relying solely on off-the-shelf GPUs, and Marvell has positioned itself as a go-to partner for designing them. That opportunity took a major step forward recently, when Marvell struck a major custom chip deal with Alphabet (NASDAQ: GOOGL) that included warrants tied to $120 billion in potential purchases.

The market read that agreement as a direct challenge to Broadcom (NASDAQ: AVGO), long the dominant force in custom AI chips, and Broadcom shares fell more than 10% on the news. For Marvell, the deal is validation that its custom silicon ambitions are real and are scaling, and it meaningfully expands the company's potential revenue per AI rack. Any commentary from management on Thursday regarding the pipeline for future custom-chip wins will be closely watched.

Marvell’s Rally Leaves Little Room for Error

Here is the counterweight every investor has to consider. Marvell stock has been on a staggering run, up 188% year-to-date, and has climbed nearly 30% over the past month. That surge has left the valuation stretched, with shares trading at a trailing price-to-earnings (P/E) near 84 and roughly 25 times sales, multiples that leave little margin for error. When a stock is priced this richly, even a strong quarter can trigger a sell-off if guidance merely meets expectations rather than blowing past them.

That dynamic has played out repeatedly this earnings season, with high-flying names beating estimates only to fall on anything short of a flawless outlook. Marvell faces the same high bar, with a great deal of optimism already priced in.

Marvell’s Earnings Will Test the Broader AI Trade

Three things will matter most. First, data center revenue and guidance: the clearest read on whether hyperscaler demand is still accelerating, as NVIDIA's outlook implies. Second, any detail on the custom silicon pipeline, particularly follow-through on the Alphabet momentum and the competitive battle with Broadcom. Third, forward guidance, since after an almost 188% run, the outlook matters far more than the backward-looking numbers.

Heading into the report, the analyst community remains constructive but not euphoric. Marvell carries a Moderate Buy consensus from 39 analysts, with an average price target of $258.42, which is only about 5% above the current price, though the most bullish target on the Street reaches $400.

For the broader market, Marvell's report carries weight beyond the company itself. NVIDIA proved the demand at the top of the stack is still surging. Marvell will test whether that strength is flowing through to the networking, interconnect, and custom silicon layers beneath it.

If data center growth and custom silicon momentum impress, it will bolster the case that the AI trade continues to broaden into the picks-and-shovels names. If guidance disappoints, it will feed the worry that expectations across AI infrastructure have simply run too far, too fast.

Read this article online ›