I went into Marvell’s Investor Day still not completely sold on the stock. Great AI exposure, sure, but at this valuation I kept wondering how much of the story was already priced in.
After yesterday... yeah, I think I was underestimating how big this could get.
* FY28 revenue guidance is now \~$20B, ahead of the Street at roughly $18.2B.
* Then they casually dropped a $70–90B FY31 revenue target. For context, Marvell did only \~$8.2B in FY26. Even getting anywhere near the bottom of that range would completely change what this company looks like.
* Custom silicon keeps getting bigger. Marvell now sees \~$12B of custom-chip revenue by FY29, up from the previous \~$10B target.
* The Google relationship might be the craziest part. The deal could represent as much as \~$120B of revenue through 2033 if Marvell hits the required milestones. Obviously that's an upside framework, not guaranteed revenue, but even a fraction of that is enormous relative to where MRVL is today.
* And I think people focusing only on custom ASICs are missing half the story. Marvell is sitting right in the middle of the AI interconnect problem: optical DSPs, silicon photonics, scale-up networking, PCIe/CXL, SerDes, XPU attach, etc. Optical interconnect had already compounded at roughly 50% annually for five years and represented around half of data-center revenue before this Investor Day.
The underlying business is already accelerating too. Last quarter revenue hit a record $2.74B, +37% YoY, while data-center revenue grew 46%, and management said AI bookings remained exceptionally strong.
Obviously the valuation is not remotely cheap (very anti value investing, my apologies), and $70–90B of revenue is an absolutely massive target that leaves plenty of room for execution failure.
But if Marvell executes even reasonably close to what they laid out yesterday, I think we're going to look back at the current business and realize we were valuing the wrong company.
I own $MRVL, so clearly biased here.
For people who watched Investor Day, what part of the long-term targets do you think is actually too aggressive? And does anyone else see MRVL increasingly becoming the closest thing to an AVGO-style custom silicon + connectivity platform, or is that taking the comparison too far?