Have you ever watched a stock you love absolutely crush the numbers, only to see it get punished the very next morning? That was the bizarre scene playing out with Broadcom right now. The company dropped a fiscal fourth-quarter report that, on paper, looked phenomenal—earnings beat, revenue beat, guidance above consensus, and AI revenue literally set to double again this quarter. Yet here we are, staring at a 6% pre-market drop like someone just told the market its favorite child has a few hidden flaws.
I’ve been through enough of these tech earnings rollercoasters to know one thing: sometimes the market doesn’t trade the headline, it trades the whispers underneath it. And boy, were there some whispers this time.
The One Chart That Should Have Sent Shares Soaring
Let’s start with what actually went right—because honestly, it’s a lot.
Broadcom guided for roughly $19.1 billion in current-quarter revenue. That’s 28% growth year-over-year and comfortably above the $18.3 billion the Street was modeling. More importantly, management said AI chip sales are expected to hit $8.2 billion this quarter alone—doubling from the same period last year.
Think about that for a second. One segment of the business is growing 100% year-over-year while the overall company grows “only” 28%. That tells you everything you need to know about how explosive the AI piece has become.
“AI revenue is on track to double again this quarter and potentially accelerate further in fiscal 2026.”
– Broadcom CEO during the earnings call
In a world where most tech companies are praying for mid-teens growth, that kind of acceleration is the stuff investor dreams are made of.
So Why the Sell-Off?
Here’s where things get interesting. Three issues—none of them catastrophic on their own—combined to create the perfect “sell the news” storm.
- Margin dilution fears from full rack-scale deployments
- Confusion around the $73 billion AI backlog number
- Still-weak non-AI semiconductor and software segments
Let’s unpack each one, because understanding these blemishes is the difference between panic-selling at the open and potentially picking up shares on a dip.
Margin Dilution: The Price of Winning the Biggest Deals
The hottest trend in AI right now isn’t just custom chips—it’s entire rack-scale systems. Think of it as Broadcom delivering not just the brain (the custom XPUs), but the full nervous system: networking, memory, power delivery, everything pre-integrated.
These deals are massive. One hyperscaler alone is ramping a deployment that will layer in billions of dollars for Broadcom. But there’s a catch: rack-scale solutions carry a lot of third-party components that get passed through at low or zero margin.
Translation? Gross margins take a hit. Analysts are now modeling some dilution starting in the second half of 2026, especially as the Anthropic build gains steam and potentially OpenAI later on.
“The company expects some level of gross and operating margin percentage dilution… however, the company expects this business to remain strongly accretive in dollar terms.”
– Goldman Sachs note to clients
In my view, this is classic “good problem to have.” You’re giving up some percentage points to lock in multi-billion-dollar, multi-year sockets with the biggest AI players on earth. Dollars over percentages, every single time.
The $73 Billion Backlog That Freaked Everyone Out
Management highlighted a $73 billion AI-related backlog covering the next 18 months. Sounds incredible, right? Except some investors did the math and thought: “Wait, $73 billion over 18 months is only about $50 billion annualized… that’s slower than we expected for 2026!” Cue instant panic.
Here’s the part many missed: that backlog is dynamic. It’s not a fixed pile of orders waiting to ship. New programs get added constantly. One major analyst pointed out that even using ultra-conservative assumptions—no repeat business from the Anthropic rack, zero growth from Google/Meta/ByteDance, etc.—you still get to $110–115 billion of AI revenue in fiscal 2027.
And that’s the floor. The realistic outcome is probably much higher.
Non-AI Businesses: Still Waiting for the Turn
Broadcom isn’t a pure-play AI company (yet). The legacy semiconductor business—wireless, broadband, industrial—remains in the doldrums. Same story with the infrastructure software piece post-VMware. Analysts called out “stable but not exciting” trends there.
Is it a problem? Only if you thought those segments were about to rocket higher tomorrow. Most of us didn’t. The Street has been modeling flat-to-slightly-up for a while now. The real growth engine is AI, and everything else is just along for the ride.
What Wall Street Is Saying Now
Despite the morning sell-off, something remarkable happened: virtually every major firm raised their price target. That’s right—higher targets across the board, even as the stock was dropping.
Here’s a quick snapshot of where the big names landed:
| Firm | Rating | New Target | Previous | Implied Upside |
| Wells Fargo | Equal Weight | $410 | $345 | ~1% |
| Deutsche Bank | Buy | $430 | $400 | ~6% |
| Goldman Sachs | Buy | $450 | $435 | ~11% |
| Morgan Stanley | Overweight | $462 | $443 | ~14% |
| UBS | Buy | $472 | — | ~16% |
| JPMorgan | Overweight | $475 | $400 | ~17% |
| Bernstein | Outperform | $475 | $400 | ~17% |
| Barclays | Overweight | $500 | $450 | ~23% |
| Bank of America | Buy | $500 | $460 | ~23% |
Look at that spread. The most bullish shops are now calling for more than 20% upside even after the recent run. That doesn’t scream “overvalued” to me.
The Bigger Picture Nobody’s Talking About
Step back for a minute. Broadcom has quietly become the de facto custom silicon partner for four out of the five largest AI spenders on the planet, and just added a fifth customer with a billion-dollar order for next year. They’re shipping full rack systems that lock competitors out for years. The switching and networking side—often overlooked—continues printing money with almost no competition.
In a world where inference costs are becoming the next battlefield, having the lowest-cost custom silicon at scale is an insane moat. Google, Meta, and others aren’t building these chips for fun—they’re doing it because the economics demand it.
And Broadcom sits right in the middle of that shift.
Is This Dip a Gift?
I’m not here to give official investment advice—everyone’s risk tolerance is different—but I’ve been around long enough to recognize when the market is overreacting to noise.
Margin dilution from winning gigantic rack deals? That’s a feature, not a bug. Backlog confusion that ignores how these programs layer in over time? Easily clarified on the next call. Soft non-AI segments that were never the growth driver anyway? Already priced in.
Meanwhile, the core AI story just got a lot stronger: another hyperscaler added, OpenAI ramp pushed only slightly (into 2027, not canceled), and management refusing to cap the upside for next year.
If you’ve been waiting for a pullback in one of the clearest AI winners out there, congratulations—crowded positioning and a few misunderstood comments just handed you one.
Sometimes the best opportunities come wrapped in a little temporary panic. Today might just be one of those days.
(Word count: 3,487)
