Rosenblatt Names Marvell a Core Long Idea, Raises PT

Rosenblatt initiated Marvell Technology (MRVL) with a Buy rating and lifted its price target to $300 from $240, calling the stock a “core long idea” ahead of Thursday’s Q2 earnings. The firm expects a beat-and-raise, driven by optical interconnect sequential growth exceeding 25% and argues FY29 earnings power of about $10.50 (29x FY29 earnings multiple), with upside from scale-up and custom-silicon ramps tied to major cloud AI platforms. While the broader Nasdaq is described as up 1.5% on Nvidia’s stellar results, the direct driver for Marvell is the raised target and earnings upside setup for the upcoming print.

Analysis

The market is still pricing AI as a compute story, but this note is really about the networking and custom-silicon bottleneck becoming the next scarce layer. That is structurally favorable for MRVL because the revenue pool tied to optical interconnect and custom ASICs should grow faster than headline cloud capex, while the company’s reuse of IP can widen incremental margins as designs roll from one hyperscaler socket to the next. The second-order winner set includes the hyperscalers themselves: AMZN and GOOGL should benefit from lower unit compute cost and better capex efficiency if their internal silicon ramps work, even if the effect is more margin-accretive than revenue-accretive.

The main loser is not necessarily NVDA in absolute terms, but the mix of spending that would otherwise have gone to merchant accelerators if custom programs scale as advertised. If the market starts to believe multi-rack AI architectures require more networking dollars per deployed compute dollar, the relative performance gap can widen between infrastructure enablers and pure accelerator names. That said, the thesis is vulnerable if hyperscaler budgets decelerate or if Marvell’s customer concentration turns into pricing pressure; those risks matter more over 1-3 months than over the next print.

Contrarian view: the stock may already be partially discounting a smooth FY28/FY29 ramp, so a beat alone may not be enough if management does not expand the long-range framework at the October event. The cleaner tell is whether order visibility and attach rates improve into calendar Q4; absent that, this can become a sell-the-news setup after an earnings pop. For the broader complex, optical suppliers and custom-silicon peers should continue to outperform only if the capex mix stays network-heavy rather than shifting back to pure compute.

Key Decisions for Investors

• Long MRVL into Thursday’s print / hold through October Investor Day: best risk/reward is if management confirms the custom-silicon and interconnect slope. Falsify on a guide that is merely in-line or any pushout in FY28/FY29 commentary.
• Pair trade: long MRVL / short NVDA for the next 1-3 months as a mix shift bet from accelerator spend to networking and custom ASIC content. Cover if NVDA re-accelerates order growth or if Marvell fails to show incremental margin leverage.
• Add AMZN or GOOGL on weakness as a lower-beta beneficiary of cheaper internal compute. This is a months-long thesis, not a one-day trade; it breaks if hyperscaler capex guides down or internal silicon adoption underperforms.
• Use MRVL post-earnings strength to fade into October if the stock gaps up on a beat without a framework raise. That is the highest-probability contrarian setup: upside likely already embedded, while the next leg requires explicit long-range visibility.