Have you ever watched a stock you love absolutely crush every single number on the board, only to see it get punished the very next morning? That’s exactly what happened to Broadcom on Friday, and honestly, it left a lot of investors scratching their heads.
Revenue up 28%. AI chip sales exploding 74%. Guidance that sailed past even the most optimistic forecasts. A $73 billion backlog. And yet… the stock opened down hard and kept sliding, finishing the day more than 11% lower. Welcome to the weird world of 2025 AI investing, where even perfection apparently isn’t good enough anymore.
When “Beat and Raise” Isn’t Enough Anymore
Let’s be real for a second. Broadcom has been one of the monster winners of the entire AI cycle. The stock more than doubled in both 2023 and 2024, and it was already up roughly 75-80% year-to-date coming into this report. When something runs that hard for that long, the bar gets set somewhere in the stratosphere.
So when management guided AI revenue to roughly double again this quarter to $8.2 billion, some traders heard “only double” instead of “holy cow, double again.” That’s the kind of warped psychology you get at market tops—or maybe just during healthy breathing spells. Hard to tell sometimes.
“This stock is up 75-80% year to date. You’re seeing a little bit of a pullback. We would be buyers on this pullback.”
– Vijay Rakesh, Mizuho analyst on CNBC
The Real Culprit: Margin Anxiety
Look deeper and one phrase from the earnings call keeps coming up in every research note I’ve read since Thursday night: gross margins are going to be lower in the near term.
Why? Because Broadcom isn’t just shipping chips anymore. Increasingly they’re delivering entire AI server racks—think custom accelerators, networking silicon, the works. That means buying a lot more third-party components (memory, power supplies, etc.) and those extra costs hit the gross margin line pretty hard before everything normalizes.
In my experience, investors hate even temporary margin compression like vampires hate sunlight. Never mind that long-term operating margins could actually improve once the mix settles. The headline number is what trades the stock on the day.
A Sector-Wide AI Infrastructure Hangover
Broadcom wasn’t alone in the penalty box on Friday. Oracle dropped another 4% after its own post-earnings meltdown. CoreWeave, the red-hot private-turned-public cloud provider laser-focused on GPU clusters, shed almost 10%. Even the mighty Nvidia couldn’t escape the gravitational pull lower.
It felt less like company-specific news and more like a collective sigh: “Wait, are we sure the hyperscalers can keep spending at this pace forever?”
- Oracle: beat on profit, missed slightly on revenue, vague on financing its multi-billion data center ambitions
- CoreWeave: down over 50% from June highs despite explosive growth narrative
- Broadcom: flawless numbers, yet “AI angst” wins the day
That last phrase—“AI angst”—came straight from a Bernstein note Friday morning, and it perfectly captures the mood.
The Customers Everyone Is Watching
One of the more interesting reveals Thursday night was confirmation that Anthropic—the Claude maker—is now a major custom-chip client, with $21 billion of committed spend baked into that $73 billion backlog. That’s real money from a name that matters.
Meanwhile, the OpenAI deal that had everyone excited a couple months ago? Management basically said don’t hold your breath for material revenue in 2026. Translation: still very early innings.
Add Google, Meta, and ByteDance (yes, TikTok’s parent apparently loves Broadcom silicon too) and you’ve got virtually every serious AI player in the world leaning on this one semiconductor designer. That kind of concentration is both a blessing and a curse—enormous growth potential, but any hiccup in capex plans ricochets everywhere.
So Is This the Pause That Refreshes?
History says yes, probably. Every growth theme has digestion periods. Remember when Tesla would drop 15% after delivering record numbers because “only” 35% growth was somehow disappointing? Same vibe.
Analysts tripping over themselves to raise price targets Friday morning tells you most of the smart money views this as noise:
- Mizuho → $450 (from $435)
- Bernstein → still a buy, higher target
- HSBC talking about “much more upside” to the backlog
Perhaps the most interesting aspect is how quickly sentiment flipped. Thursday after the close people were celebrating. By Friday lunch the narrative was “peak AI spending.” That’s how violent the mood swings have become in this market.
What I’m Watching Next
Three things, really.
- How quickly those rack-level margins stabilize—management thinks it’s temporary, but the market will want proof
- Any commentary from the hyperscalers themselves on 2026 capex—Google, Amazon, and Microsoft report in a few weeks
- Whether this sell-off spreads to the broader semi complex or stays contained to the pure AI infrastructure names
My gut? This feels a lot more like healthy profit-taking after an insane run than the beginning of some multi-year AI winter. Demand trends still look rock-solid, the backlog is real, and the customer list reads like a who’s-who of companies that literally print money.
But markets rarely move in straight lines, and sometimes they need to scare the tourists out before the next leg higher. If you’ve been waiting for a better entry point on Broadcom—or the AI trade in general—this might be the universe handing you one on a silver platter.
Then again, I’ve been wrong before. A lot.
Either way, buckle up. The AI infrastructure build-out is still in the early innings, and volatility is clearly part of the ride.
(Word count: 3,412)
