Jim Cramer Says Marvell Is Expensive Unless Everything Works, and Then It Is Cheap

Custom Silicon in Plain Language

Custom silicon, or ASIC, work means designing a chip for a specific customer’s workload rather than for general use. A general-purpose GPU like NVIDIA’s is programmable across almost every AI model. A custom XPU is built to the hyperscaler’s spec and generally will not be resold elsewhere.

Marvell’s expanded commercial agreement with a large hyperscaler covers custom programs already in execution, awards made over the past several years, new design wins, and potential future programs, and the company disclosed a warrant allowing Google to acquire up to 7% of Marvell’s shares, tied to revenue milestones. Matt Murphy said the deal reflects “the scale and long-term potential of the relationship”. This structure aligns the customer with Marvell’s success and concentrates the outcome.

Murphy told analysts that Marvell expects the custom business to more than double year over year in fiscal 2028 and accelerate significantly in fiscal 2029. Note that Marvell reports on a fiscal calendar, so fiscal 2028 runs through early calendar 2028.

NVIDIA’s Contrast

NVIDIA (NASDAQ:NVDA) reported total revenue of $96 billion, more than doubled year over year, and management guided to approximately 70% growth in fiscal 2028, which Jensen Huang called “a supply-constrained outlook”. Its trailing PE is 32 against a 63% profit margin.

NVIDIA is priced for a company that already dominates, while Marvell is priced for one that still has to prove the second act (we reverse-engineered what the biggest tech winners looked like early and put the pattern in a free playbook here: The Next Nvidia Playbook).

Marvell’s operating margin is 14.5% on a trailing basis, and management expects to move into the 38%-40% non-GAAP operating margin range as fiscal 2028 progresses. The gap between reported profitability and target profitability is where the valuation risk lives.

Is October 6 a Real Catalyst

Cramer’s second remark was that you do not want to bet against the stock going into the October 6 meeting because Murphy is going to tell a very compelling story. He also noted the stock did not take out the top, so traders will call it a double top.

Both can be true. The analyst meeting is a genuine catalyst because management has explicitly promised a detailed fiscal 2029 and beyond revenue framework, ranges of potential outcomes for the expanded hyperscaler agreement, and an updated long-term operating model. Murphy said investors should assume that on the custom side, “these numbers would be a lot larger than anybody’s been modeling so far”.

The thesis fails in one specific way: customer concentration. Marvell disclosed dependence on a few customers for a significant portion of its revenue, with data center now accounting for 79% of total revenue. If a top hyperscaler pulls a program in-house, the same custom win that drives the bull case becomes the concentration risk that unwinds it. The Q2 FY27 8-K lays out that risk in the disclosures at sec.gov. The stock likely trades in a range into October 6 unless Murphy delivers a fiscal 2029 number that closes the gap between price and proof.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Marvell Technology didn’t make the cut. Grab the names FREE today.

Contacteditorial@247wallst.com for any questions or corrections.