Where we heard itCNBC’s Squawk on the Street, Oct. 7, 2026. Speaker: Marvell Technology chairman and CEO Matt Murphy, interviewed by Jim Cramer and David Faber. Watch at CNBC
Marvell TechnologyMRVL closed at $274.66 on October 8, 2026, down 3.5% on the day. The stock is up 223% year to date and 197% over twelve months.
A stock up this much on the year is defending a re-rating already priced in. That background frames the long-range targets Marvell laid out at its October 6, 2026 Investor Day and CEO Matt Murphy’s comments the following day.
Investor Day Targets Reach Out to 2031
Marvell raised its 2028 revenue forecast on strong AI data center demand, setting a fiscal 2028 revenue target of $20 billion, according to Reuters.
The longer target extends further out. Marvell set a fiscal 2031 revenue goal of $70 billion to $90 billion at the same Investor Day.
Murphy Calls a $30 Billion Custom Chip Target “Not a Stretch”
On October 7, CEO Matt Murphy told CNBC a $30B custom chip target is “not a stretch.” Jim Cramer also raised Nvidia CEO Jensen Huang’s June prediction that Marvell could become a $1 trillion company. Murphy called that remark “not rehearsed and shocking,” then said that if you take the $70 billion to $90 billion range and put a multiple on it, four years out, “who knows what’s possible.” That remark is separate from the company’s published targets.
Murphy’s Case Rests on the Connectivity Layer
Murphy’s reasoning centers on the AI infrastructure bottleneck. In the interview, he described the cycle: “We’re unique actually in this AI semiconductor cycle in that you had the three big compute companies, right, drive massive growth when generative AI first rolled out. That was followed by a huge memory wave … And what we’re seeing now is there’s a gap because there’s not enough in the connectivity layer to connect all of those things.”
He continued: “You’ve got a lot of installed compute, a lot of installed memory, but guess what? They all need to talk to each other through high-speed optical interconnections, switching networks. And what you don’t want is any GPUs or memory sitting idle.”
In short, Murphy argues the limited resource in AI infrastructure has changed to the plumbing between processors and memory. Idle accelerators are expensive hardware producing nothing. So the links and switches that keep them busy gain strategic value.
How a 13% Share Framing Shapes the Pitch
Murphy framed the custom target against market size, noting the custom silicon market is $235 billion by 2030 and Marvell is targeting $30 billion in custom revenue that year. He noted this equals “only 13% market share.” The $235 billion market size is Marvell’s own estimate, presented at its Investor Day, and has not been independently verified.
Marvell’s targets at a glance
Framing a large revenue goal as a small slice of a much larger market makes it sound more attainable, and investors can weigh the market size estimate separately from the share claim.
Marvell’s own disclosures list risks that bear on both: dependence on a few large customers, heavy concentration in the data center end market, and the possibility that customers develop their own solutions or vertically integrate. The full risk factors are available through Marvell’s investor relations site.
Long Runway for Proof, Short Path for the Stock
The targets run out to fiscal 2031, which for Marvell ends in early 2031. Murphy described the custom silicon goal as a 2030 figure.
The stock has already moved sharply, up 223% this year and 325% over five years. That run-up suggests investors have already credited much of the connectivity thesis, and tracking how each reported period compares with the Investor Day targets will test Murphy’s argument.
Murphy argues the next AI bottleneck is the connection between processors and memory, and that Marvell’s $30 billion custom chip goal needs only about 13% of a $235 billion market. The risk is that a stock up 223% this year already prices in much of that, while Marvell leans on a few large customers that could build more of their own chips.