If Broadcom can grow AI semiconductor revenue 143% year over year and still lose more than $250 billion in market value, what level of growth is now “enough” for the AI trade?
That is the real question behind the selloff. Broadcom reported fiscal second-quarter 2026 results on June 3, with AI semiconductor revenue reaching $10.8 billion, but its third-quarter AI chip forecast landed below Wall Street’s expectations, according to CryptoBriefing. Shares dropped more than 13% in after-hours trading, and in some reports the decline reached 13% to 15%.
This was not a collapse in the business. It was a collision between strong results and stronger expectations.
How did 143% AI revenue growth turn into a $250 billion punishment?
Broadcom’s quarter exposed a harsher phase of the AI infrastructure trade: investors are no longer paying up for evidence that demand exists. They want proof that demand is accelerating faster than already aggressive models assume.
The company reported overall fiscal second-quarter revenue of $22.19 billion. Its AI semiconductor segment delivered $10.8 billion, slightly above Broadcom’s internal guidance. On the surface, that is the kind of print investors used to reward.
The problem was the next number. Broadcom guided for $16 billion in third-quarter AI chip revenue. Wall Street consensus estimates ranged from roughly $16.36 billion to $17.2 billion. That left the guide about $360 million to $1.2 billion short, depending on the estimate used.
The gap was small relative to the forecast. The stock reaction was not.
| Metric | Reported / guided figure | Market context |
|---|---|---|
| Q2 AI semiconductor revenue | $10.8 billion | Up 143% year over year |
| Q3 AI chip revenue forecast | $16 billion | Roughly 48% above Q2 AI revenue |
| Wall Street expectation range | $16.36 billion to $17.2 billion | Forecast missed the low end |
| After-hours share move | Down more than 13% | Over $250 billion in market value erased |
MLXIO analysis: the market was not rejecting Broadcom’s AI business. It was repricing the certainty investors had attached to that business. A 2% to 7% forecast shortfall against analyst expectations triggered a double-digit stock move because future AI growth had become central to the equity story.
Why was the guidance more important than the quarter Broadcom just reported?
Because Broadcom has become a proxy for whether hyperscaler AI infrastructure spending is still ramping cleanly.
CEO Hock Tan highlighted demand from Anthropic, Google, Meta, and OpenAI. He also reiterated Broadcom’s long-term target of surpassing $100 billion in AI semiconductor revenue by 2027, which implies roughly $25 billion per quarter if sustained evenly.
That target did not move higher. For a stock priced around aggressive AI expectations, “unchanged” can read like disappointment.
Reuters reporting published by The Straits Times quoted Ben Bajarin, CEO of Creative Strategies, on the issue:
“Nothing slows down what was estimated prior – they just didn’t raise it,”
That sentence captures the market’s mood. Broadcom did not guide to weakness. It simply failed to stretch the story further.
The company also disclosed plans to ship more than 10 gigawatts of AI compute capacity in 2027, with growth expected to continue into 2028. Tan said Broadcom planned “a lot more” in 2028. Executives also said Broadcom was “very comfortable” it had secured supply for 2026 and 2027.
Those are not soft signals. But the stock reaction says investors wanted a cleaner upward revision.
What does Broadcom’s custom AI chip role reveal about hyperscaler spending?
Broadcom matters because it is not just another AI-adjacent semiconductor name. It supplies custom AI accelerators, AI networking, and related infrastructure components for large cloud and internet companies.
That role puts Broadcom in a different lane from Nvidia, whose GPUs remain the industry standard for many AI workloads. Broadcom’s opportunity sits in the custom silicon push: hyperscalers building or commissioning chips tailored to their own workloads, cost structures, and deployment plans.
The source material also points to competition from Marvell Technology, which said at the end of May that its custom chip business would surpass $10 billion in revenue in 2029 and forecast second-quarter revenue above estimates.
That matters because custom AI silicon is strategic but not necessarily smooth. Orders can be large, concentrated, and tied to deployment schedules at a small number of hyperscale customers. If one buildout shifts, revenue visibility can change quickly.
For readers tracking how AI silicon choices show up beyond data centers, MLXIO has also covered device-level hardware pressure in 45 TOPS Bet Turns Asus Zenbook 14 Into AI Chip War and 1.4kg Dell Precision 14S Forces a Brutal Chip Gamble. Those are separate product stories, but they reflect the same broader point: silicon decisions now carry more commercial weight than branding alone.
Why does this feel different from the early AI boom?
The early AI trade rewarded exposure. A company only needed credible AI demand to attract capital. Broadcom’s reaction suggests that phase is fading.
Now, investors are testing three harder variables:
- Visibility: Are orders recurring or front-loaded?
- Timing: Do hyperscaler deployments arrive in steady ramps or uneven waves?
- Execution: Can suppliers convert demand into shipped capacity without bottlenecks?
Broadcom’s case is notable because the company is not a fragile pure-play. It posted $22.19 billion in quarterly revenue and has a large non-AI business. Its AI revenue is growing rapidly. Its customer list includes some of the most important AI buyers in the market.
Yet the stock still sold off sharply.
MLXIO analysis: that tells investors something uncomfortable. AI infrastructure equities may now trade less on the existence of demand and more on whether management can beat the most optimistic version of that demand every quarter.
Are bulls seeing a pause while bears see an expectations bubble?
The bullish case is straightforward. Broadcom remains tied to custom accelerators, AI networking, hyperscaler demand, and a long-term target above $100 billion in AI semiconductor revenue by 2027. The Q3 AI forecast of $16 billion still implies a steep sequential increase from $10.8 billion.
Tan’s statement also supports the demand view:
“Q2 semiconductor revenue from AI of US$10.8 billion grew 143 per cent year over year, above our forecast, driven by increasing demand for custom AI accelerators and AI networking,”
The bearish case is not that Broadcom’s AI business is broken. It is that the stock had little room for anything less than upside. Ryan Lee, senior vice-president of product and strategy at Direxion, put it bluntly:
“Today’s miss on revenue and subsequent post-market pullback (in shares) shows the market demands perfection for this chip rally to keep running,”
That is the key tension. Bulls can point to real growth. Bears can point to valuation sensitivity. Both can be right at the same time.
What should AI infrastructure investors take from Broadcom’s miss?
The useful lesson is not “AI spending is over.” The source material does not support that. Big Tech firms are expected to spend more than US$700 billion on AI infrastructure in 2026, up from around US$400 billion in 2025, according to Reuters reporting cited in The Straits Times.
The lesson is that large AI spending does not automatically protect every AI-linked stock from guidance risk.
Semiconductor peers, networking suppliers, memory makers, and cloud infrastructure vendors exposed to AI capex may face the same question Broadcom just faced: are their forecasts rising fast enough to justify the growth already priced in?
Enterprise AI adoption also sits in the background. If returns on AI spending take longer to show up, infrastructure suppliers could face sharper scrutiny over order timing and customer concentration. That is an inference, not a reported fact. But it follows from the market reaction to a forecast miss that was numerically modest and financially brutal.
What evidence would prove Broadcom can outrun valuation pressure?
Broadcom’s next test is not just another strong quarter. It needs stronger guideposts.
The evidence that would support the bullish thesis is clear: AI chip revenue tracking toward the $25 billion-per-quarter pace implied by the 2027 target, repeat demand from named hyperscalers, and continued confidence around supply for 2026 and 2027.
The evidence that would weaken it is equally clear: another guide below analyst expectations, slower progress toward the $100 billion AI semiconductor goal, or signs that custom silicon orders are arriving in bursts rather than a steady ramp.
Broadcom’s selloff shows the AI infrastructure trade is maturing. Narrative still matters, but numbers now matter more. The market is no longer asking whether AI demand is real. It is asking whether the next forecast can beat the last assumption.
Disclaimer: This MLXIO analysis is for informational and educational purposes only. It is not financial, investment, legal, tax, or professional advice. It does not provide buy, sell, hold, price-target, portfolio, or personalized recommendations. Verify information independently and consult qualified professionals before making decisions.
The Bottom Line
- Broadcom’s selloff shows AI stocks are being judged against extremely high growth expectations.
- Even 143% year-over-year AI revenue growth was not enough when guidance missed consensus.
- The reaction signals investors may be demanding faster acceleration across the AI infrastructure trade.
