Let me start with Broadcom, because that’s where the reaction originated.

Broadcom has been around in its current form since the 2016 merger of Avago Technologies and the original Broadcom. Most investors know the name but fewer know how wide the actual business runs. Semiconductors, yes, but also a sprawling infrastructure software segment that got significantly larger after the VMware acquisition closed in late 2023. That software layer is what makes AVGO different from a pure chip play. It adds recurring revenue, smooths out cyclicality, and gives Hock Tan a lever to pull that most of his peers simply do not have.

Tan has been CEO since 2006. His style is not flashy. He acquires, cuts aggressively, runs lean operations, and sets big long-range targets in public. That last habit is what created this week’s problem.

Fiscal Q2 2026 results, reported June 3: revenue of $22.187B, up 48% year over year. Non GAAP diluted EPS of $2.44. AI semiconductor revenue specifically hit $10.8B for the quarter, up 143% year over year. Q3 guidance came in at roughly $29.4B, representing 84% year over year growth. Those are not weak numbers. They are genuinely strong numbers.

And yet.

Management kept the fiscal 2027 AI semiconductor sales target at more than $100B instead of raising it. That single decision is what the market fixated on. Not the quarter, not the beat, not the guidance step-up into Q3. The long-range number stayed flat and that read as a signal, fair or not, that the ceiling had been found. Shares dropped hard. The reaction felt disproportionate to me, but the market is not always in the business of being proportionate.

On the innovation side, what Broadcom has built is harder to replicate than it looks. The custom accelerator programs with hyperscalers are deep, multi-year co-development relationships. You do not swap out a chip partner mid-cluster build. The AI networking silicon that wraps around those accelerators adds another layer of stickiness. This is not a commodity exposure to AI spending. It is closer to infrastructure you build on top of.

Slight tangent, but it matters: the VMware integration is still in progress. A lot of investor attention goes to the AI chip side and almost none goes to the software segment. That asymmetry might be worth thinking about if you are building a longer-term view on AVGO.

Now Marvell.

Marvell Technology is a different animal. It has spent the last several years repositioning almost entirely around data center infrastructure, shedding consumer and storage exposure to focus on the parts of the market that AI spending hits hardest. The Inphi acquisition, completed in 2021, was the move that really shifted the portfolio. Inphi brought electro optics and high-speed interconnect capability that now sits at the center of how hyperscalers move data between chips and between racks. Matt Murphy, Chairman and CEO since 2016, has been deliberate about this transformation in a way that did not always get credit from the market until the AI spending wave made it obvious.

Fiscal Q1 2027 results, reported June 3: record revenue of $2.418B, up 28% year over year. Non GAAP diluted EPS of $0.80. For Q2, Marvell guided revenue to $2.7B at the midpoint and non GAAP EPS to $0.93 plus or minus five cents. Street estimates heading into that guidance were sitting around $2.60B in revenue and $0.90 in EPS. Marvell beat both and then guided above both. That is the version of this week that Broadcom did not deliver.

What’s interesting is the contrast in how the two companies communicated. Murphy was direct about bookings strength and the multi-year demand picture. Optical DSP platforms for faster data center interconnects, custom silicon programs for hyperscalers, AI networking demand that is accelerating rather than flattening. The language matched the numbers and then went further than the numbers. That combination is what the market rewarded. Broadcom’s AI revenue was actually larger in absolute terms, but the framing around it was more cautious.

Here is where I land on both of these: the underlying demand is real for each company. What shifted this week is the market’s willingness to bid up stocks where management is not explicitly raising the ceiling. That may or may not be rational depending on your time horizon. If you own AVGO, the quarter did not break the thesis. If you are watching MRVL, the guidance beat matters more than people are giving it credit for, because Murphy went out of his way to be specific and optimistic at the same time.

The thing I keep coming back to is this: both companies are building real infrastructure for a real spending cycle. The gap between their stock reactions this week was almost entirely about how their CEOs chose to talk about the next 18 months. That is a strange variable to hang this much volatility on.

Full breakdown here if you want to go deeper on either name before next earnings.