Broadcom Fell 5% After Beating Every Metric. Here's Why That Matters.

Four AI infrastructure companies reported earnings in a 48-hour window last week. Three of them beat Wall Street's numbers. One of the three still fell 5%.

That one was Broadcom.

Broadcom reported $29.6 billion in quarterly revenue—up 86% from a year ago. Its AI semiconductor business generated $16.7 billion, up 221%, and it guided for $21.7 billion next quarter, which is 236% year-over-year growth. The stock dropped about 5% after the bell because its Q4 revenue guidance of roughly $34.8 billion missed Wall Street's consensus estimate of $35.03 billion.

Meanwhile, Snowflake reported adjusted EPS of $0.62 versus a consensus of $0.26—more than double what analysts expected—and its product revenue grew 37% year-over-year to $1.49 billion. Shares jumped 22%. Hewlett Packard Enterprise posted $12.2 billion in revenue against $11.9 billion expected and $1.11 in adjusted EPS versus $0.95 consensus. The stock rose roughly 2%. Dell surged about 10% after raising its AI server revenue outlook.

The headline across all four names is the same: AI demand remains strong. The market's reaction across those same four names is the opposite for Broadcom. The reason Broadcom fell tells you something about how the plumbing of this trade has changed.

The tolerance gap

When Broadcom guided for $34.8 billion next quarter, that figure represents 93% year-over-year growth on a scale that would make almost any company in history envious. It's also $230 million below what analysts had extrapolated. On a $35 billion quarter, that's roughly a 0.65% miss.

But in the current regime, the AI infrastructure trade has developed a precision tolerance for growth that didn't exist a year ago. Broadcom beat every single quarterly metric—revenue, earnings, AI semiconductor bookings, operating income—and the only thing that mattered was the $230 million gap between $34.8 billion and $35.03 billion. The market priced the quarter perfectly and then re-priced the stock on the guidance decimal.

Snowflake had the mirror experience. It beat EPS by 36 cents on an estimate of 26 cents—roughly a 140% surprise—and revenue growth accelerated to 37% from the prior quarter. That acceleration beat the consensus for product revenue growth of roughly 30%, and the market responded by adding 22% to the stock in after-hours trading.

Same AI boom. Same underlying demand story. Different reaction because one company's guidance landed above the consensus line and one landed below it. That is not a reflection of business fundamentals. It's a reflection of positioning and expectations that have been pushed to a point where the market no longer rewards good results—it punishes good-enough ones.

Why the tolerance tightened

This precision trading didn't appear in a vacuum. Broadcom's stock has more than doubled over the past year, and the market bid it up by extrapolating the exact same growth trajectory forward, quarter by quarter, each time analysts raised the bar. The company has historically beaten those bars by a small margin, which reinforced the behavior. Until this quarter, when it didn't.

At the same time, Broadcom's gross margins are being pressured by the shift toward AI semiconductors, particularly custom TPUs, which carry lower margins than the company's legacy portfolio. Consolidated gross margin is running around 74%, down from earlier levels. Management guided for roughly 66% non-GAAP operating margins next quarter, flat year-over-year. Revenue is growing at 93%, but margins aren't growing with it. That dynamic matters less when the stock is being rewarded for top-line acceleration and more when it's being punished for the smallest hint of deceleration.

HPE shows a different face of the same trade. The stock is up approximately 140% over the past four months and 116% year-to-date, driven by AI server demand and the integration of its $14 billion Juniper Networks acquisition. HPE beat Q3 on both revenue and EPS and guided for 34% to 37% full-year revenue growth, with networking growing 73% to 74%. The stock rose about 2% on the report—not a huge move for a company that has already run up this far.

The point is that HPE has been rewarded for hitting the beat, not for accelerating expectations. Broadcom was punished for not accelerating them. The same trade treats the two companies differently only because Broadcom was priced to perfection and HPE was priced for strong execution. The market doesn't care about the AI story. It cares about whether you beat the bar that was already built into your price.

The options layer

The positioning underneath this cluster of earnings tells part of the story too. SPY options showed an average implied volatility of 13.41%, which looks calm until you look at the structure beneath it. The put-to-call volume ratio on SPY was 1.24, meaning traders were buying roughly 24% more put protection than call exposure across the options market. The put-to-call open interest ratio was 2.46—nearly two and a half puts for every call outstanding.

That is not the options positioning of a market that's sleeping on its wins. It's the positioning of participants who know what's coming—four correlated AI infrastructure names all reporting within a day of each other—and are buying downside insurance around the event risk. The market was hedging the precision trade before it even happened.

Broadcom's own options picture reinforces the fragility. The stock carries an average implied volatility of 74%, which reflects the market's expectation that the stock will swing sharply around its earnings. The put-to-call volume ratio was 0.66, suggesting more call buying in the short term, but the put-to-call open interest ratio sat at 1.07—roughly even, meaning dealers sit with significant put gamma on the other side of those calls. If the stock moves down through a key put wall, dealer hedging can accelerate the move rather than dampen it.

This isn't abstract positioning data. It's the mechanical layer that determines how a 0.65% guidance miss becomes a 5% stock move. Dealers who sold calls into Broadcom's pre-earnings rally were short gamma and had to buy shares as the stock moved down, amplifying the decline. The earnings report provided the catalyst; the options structure provided the force multiplier.

What it means for the trade

The consensus narrative going into this earnings cluster was straightforward: AI demand is strong, the companies are delivering, the stock market should respond positively. That narrative was directionally correct—three of four names beat, and three of four stocks rose. But the narrative missed the mechanism.

The mechanism is that Broadcom's AI revenue more than tripled and the stock still fell because the guidance was $230 million shy of a number that doesn't represent any real business constraint—it represents an analyst extrapolation. The trade has moved from "will AI demand continue?" to "will it continue at exactly the rate the market has already priced in?" The answer to the first question was yes. The answer to the second is harder, because it requires a company to grow at an accelerating clip in perpetuity while maintaining margins, and Broadcom signaled that margins won't expand alongside the revenue growth.

That's the tension these earnings laid bare. The AI infrastructure buildout is real. The bookings are real. The revenue growth is real. What may not be real is the tolerance for anything less than flawless execution across the entire supply chain. Broadcom showed what happens when one link in that chain signals a fraction-percentage shortfall at $35 billion in scale. The market re-prices the stock as if the thesis is broken, even though the business is generating $16.7 billion in AI revenue in a single quarter.

The question for anyone holding or watching these names isn't whether AI demand will persist over the next three to five years. The question is whether the current pricing assumes a trajectory that no company can sustain at this scale, and whether the next 0.65% miss triggers the same kind of repricing. Broadcom answered that question with today's move. The market's response tells you the positioning is still concentrated, the tolerance is still thin, and the options structure still amplifies downside surprises.

The plumbing hasn't changed. The earnings did.