Management told us cloud capex growth will moderate to 30%+, then committed $1 billion in supplier prepayments to lock capacity against that same slowing demand.
Thesis: Marvell owns the interconnect bottleneck in the AI data center buildout. Interconnect growth went 30% to 50% to 70% in three quarters, custom revenue doubles in FY2028, and $1B in prepayments locks supply behind demand signals from a handful of hyperscalers. The operating leverage math works if cloud capex holds at 30%+ growth, but management itself projects moderation into FY2028 while committing cash ahead of confirmed orders. The next two quarters determine whether this is structural acceleration or peak cycle behavior.
Verdict: HOLD — Conviction: MEDIUM
Catalyst: Q2 data center revenue growing mid-to-high teens sequentially, validating double-digit sequential growth for all of FY2027 and pulling FY2028 estimates higher as the new Tier 1 XPU program hits volume production milestones.
Key Risk: One custom program slips or one hyperscaler normalizes capex and the FY2028 $16.5B guide breaks. The new Tier 1 XPU contributes ~1/3 of FY2028 custom growth on a 2-year design cycle with no named customer and no cancellation terms disclosed.
The Tell: Murphy volunteered 'upward bias' unprompted when asked about the $16.5B FY2028 number, then repeated it: 'there's a possibility for a lot of upward bias...I think there's upward bias for sure.' CFO simultaneously said cloud capex is moderating to '30% plus range.' CEO calling for raises while CFO warns of capex deceleration on the same call is the tell.
Friction Level: MODERATE_FRICTION — Both sides agree on clean numbers and rising prepayments. The disagreement is whether accelerating design wins and the NVIDIA partnership are structural or whether peak confidence plus self-forecasted capex moderation makes FY2028 a coin flip.
Report not found
The report data is no longer available. Please return to the archive.