Marvell (MRVL) +10% Small Q4 beat with some DC nits, but much better out year guide/commentary driven by optics/interconnect and improving DC demand has shorts scrambling a bit
Stock has been one of more hated and shorted semis among HF community and the FY27/FY28 numbers, which were better than the street and bogeys which were at or slightly below, will have some wanting to cover although still remains a show me story given mgmt’s track record.
Key quote from the call: “…expect year-over-year revenue growth to accelerate each quarter in fiscal 2027, driven by continued strength in our data center business, with bookings continuing to grow at a record pace” which will bring them to at least 30%+ growth for FY27, approaching $11B driven by data center at +40% y/y, with interconnect at more than +50% y/y. FY28 was also reset higher, with total revenue now ~$15B, data center close to +50% y/y, and non-GAAP EPS well over $5.
The #s:
Quarter mattered less than the FY27/FY28 framework but were ok:
Revenue was $2.219B, +22.1% y/y (last q +36.8% y/y) versus Street at roughly $2.21B, and EPS was $0.80 versus Street at $0.79. Data center was $1.651B, +20.9% y/y, also modestly ahead. FQ1 guide was the bigger surprise, with revenue at $2.4B midpoint, +26.6% y/y, versus Street around $2.27B to $2.28B, and EPS at $0.79 midpoint versus Street at roughly $0.75. ..mgmt also said OpEx should stay flat sequentially in FQ2 before rising only low to mid single digits in FQ3 and FQ4.
Key Takeaways:
Mgmt effectively moved the narrative from interconnect growing more like cloud capex to growing more like accelerator demand. 800G remains strong, 1.6T is now ramping much faster, and mgmt’s tone on electro-optics was much more forceful, including commentary that the business is now growing 50%+ and can keep that momentum into FY28.
Mgmt reiterated custom at more than +20% y/y in FY27 and at least doubling in FY28, while also sounding firmer on purchase orders, volume requirements, XPU attach ramps, and the next Tier 1 XPU program. The pushback point is that the exact magnitude of the big XPU ramps still matters a lot, but the tone was clearly more confident than cautious.
The scale-up optionality got stronger. Celestial AI and XConn are now closed, CPO for scale-up is still expected to start next year, AEC and retimers are expected to more than double in FY27, and switch revenue is now expected to surpass $600M in FY27.
Gets an ug at BofA saying the update increased confidence in Marvell’s leverage to AI optical connectivity, potential participation in Microsoft’s upcoming custom chip program, and signs the company is “turning the corner” following the Amazon XPU transition year.
Bull vs Bear Debate
Bulls say MRVL is becoming one of the broadest AI infrastructure plays outside the very largest names, and this quarter made that argument stronger. Bulls are not looking at MRVL as just “another custom ASIC supplier.” They see a company with real exposure across optical interconnect, PAM and coherent DSPs, 1.6T migration, AECs, retimers, switching, CXL and NIC attach, plus scale-up CPO optionality from Celestial AI and protocol/switching optionality from XConn. The quarter reinforced that the biggest upside driver right now is not legacy recovery or even only custom. It is optical/interconnect accelerating harder than expected, with mgmt explicitly moving the FY27 interconnect view to more than +50% from +30%, while also talking about stronger bookings, rising cloud capex, and demand visibility that now reaches well into FY28. Bulls also think the custom debate is getting too anchored to the noisy parts of AMZN and MSFT timing, while missing how much XPU attach and networking can do.