Skip to content
AdClaim Your Free Copy: The Weekly Options Strategy Anyone Can Use

Articles

Marvell's AI Chip Opportunity Is 5x Bigger Than It Admits: BofA

Sep 14, 2026 · Trading Tips

Marvell Technology stock has nearly tripled this year, up about 180% since January. That kind of run usually invites skepticism about how much further it can go. Bank of America's answer: a lot further.

BofA analyst Vivek Arya reiterated his Buy rating and $365 price target on Marvell (NASDAQ: MRVL) this week, implying roughly 54% upside from where the stock closed on September 11 near $236. Shares rose 4.7% Friday morning on the note.

The call came after Arya hosted Marvell's CEO Matt Murphy and CFO Dan Durn at an investor lunch in New York. Arya walked away convinced Marvell itself is underestimating how big its AI opportunity could get — and he's putting real numbers behind that view.

His math: the market for custom AI chips, plus the components wired to them, could reach roughly $300 billion by 2030. That's five to six times the $55 billion Marvell itself projected for 2028 back in June 2025 — a gap wide enough that even a modest capture rate would move the needle on earnings.

"The chips around the chip" — connectivity and memory components sitting next to every custom AI processor — are what BofA calls harder to displace than the processors themselves.

The core of the thesis is something Marvell calls "XPU attach" — the connectivity and memory chips that sit next to every custom AI processor cloud giants design in-house. Arya says these are stickier and harder to displace than the processors themselves, because they're wired into a customer's specific chip design early in the process.

Each processor can carry one or more of these attach chips, priced at $500 to $1,500 apiece — a big step up from the roughly $50 storage controllers Marvell used to sell. Arya's model assumes about 30 million custom processors ship in 2030, translating into a $60 billion to $65 billion attach market where Marvell could capture $30 billion at a 40% to 50% share.

For context, Marvell's own current guidance calls for just over $1 billion in attach sales in 2027, scaling to $3 billion to $4 billion in 2028. If Arya's numbers are even directionally right, that guidance has a lot of room to move up over the next two years.

On the processors themselves, Arya sees a $350 billion to $400 billion custom-chip market by 2030 — about a quarter of total AI chip spending. Strip out high-bandwidth memory, which Marvell doesn't sell, and there's roughly $240 billion left for Marvell to chase. A 5% to 7% share of that translates to $12 billion to $17 billion in sales, well above Marvell's current sub-$10 billion 2028 outlook, which Arya called "very conservative."

The customer picture backs this up. Amazon, Marvell's lead customer, is expected to keep growing orders every year. Microsoft is set to start ramping its own custom silicon relationship with Marvell next year. Management also said a recent Qualcomm-Amazon optics deal has zero impact on Marvell's own optics business, and that its Alphabet relationship is limited to networking and connectivity with no overlap on existing products.

On earnings power, Arya estimates Marvell could generate $14 a share by calendar 2028 if the AI attach business scales as projected — versus $11 a share if custom chip sales stay flat around $11 billion. Every extra $1 billion in sales could add 30 to 35 cents to that per-share number.

For investors sitting on Marvell gains after this year's run, the case for holding rather than trimming rests on the October 6 analyst day, where management could formally update its market sizing and margin targets. A guidance raise that starts closing the gap with Arya's numbers would give the stock another catalyst; a quiet reiteration of current targets would leave the stock trading more on sentiment.

The risk here is straightforward: Marvell is already up 180% on the year, and a lot of AI-optimism is baked into the price. Any wobble in AI capex sentiment — like the chip-sector selloff triggered by AI-slowdown concerns this week — could hit a high-multiple name like Marvell harder than the broader market.

Bottom line: BofA thinks Wall Street, and maybe even Marvell's own management, are still lowballing the AI attach opportunity — watch October 6 for confirmation.