Broadcom's AI Chip Revenue Soared 143% to a Record — and Its Stock Still Fell

Broadcom delivered the kind of quarter that, on paper, validates the entire AI investment thesis. The chipmaker reported record revenue, a 143% jump in artificial-intelligence chip sales, and guidance pointing to even faster growth ahead. Then its stock fell.

Shares of Broadcom slipped roughly 3% in after-hours trading Wednesday after the company posted fiscal second-quarter results that beat Wall Street’s profit estimate and set a revenue record. The drop wasn’t a verdict on the business, which is expanding at a pace almost no company its size can match. It was a verdict on expectations — and a reminder that, with major stock indexes sitting at record highs, even a blowout AI quarter now has to clear an extraordinarily high bar to satisfy investors.

That gap between exceptional results and a falling share price is the story worth understanding. It says less about Broadcom than about the state of the AI trade carrying the broader market.

The Numbers Behind the Record Quarter

The headline figures were strong by almost any standard. Broadcom reported consolidated revenue of $22.2 billion for the quarter ended in early May, up 48% from a year earlier and a company record. Adjusted earnings came in at $2.44 per share, edging past the roughly $2.40 analysts expected.

The engine was artificial intelligence. Broadcom said semiconductor revenue from AI reached $10.8 billion in the quarter, up 143% year over year and above its own prior forecast, driven by demand for the custom AI accelerators it designs for a small group of large cloud customers and the networking chips that tie those systems together. Adjusted EBITDA rose 52% to a record $15.2 billion, or 69% of revenue, and free cash flow hit $10.3 billion.

It is worth understanding what Broadcom actually sells, because it shapes how investors read these numbers. Unlike Nvidia, which sells merchant graphics processors to almost anyone building AI systems, Broadcom designs custom accelerators — purpose-built chips, sometimes called ASICs — for a handful of the largest cloud companies that want silicon tailored to their own workloads rather than off-the-shelf GPUs. It pairs those chips with the high-speed networking gear that stitches thousands of accelerators into a single training cluster. That model concentrates Broadcom’s AI revenue in a small set of enormous customers, which makes the trajectory steeper but also more dependent on each of those buyers continuing to spend. When a few hyperscalers commit to multiyear build-outs, the order book balloons; when any of them pauses, the effect is outsized. That concentration is part of why the market scrutinizes Broadcom’s forward guidance so intensely.

Not every line was a beat. Broadcom’s infrastructure software unit — the business built largely around its acquisition of VMware — posted revenue of about $7.18 billion, just under the roughly $7.32 billion analysts had penciled in. And the company’s total revenue, while a record, came in fractionally below the consensus estimate of about $22.27 billion. In a quarter defined by the size of the AI number, those were small misses. They were not what moved the stock.

The forward guidance was, if anything, more striking than the quarter itself. For the current quarter, Broadcom guided to revenue of about $29.4 billion, which would represent 84% growth from a year earlier, and said it expects AI semiconductor revenue to grow more than 200% year over year to roughly $16 billion. Chief Executive Hock Tan told analysts the AI demand momentum is continuing.

Why a Beat Wasn’t Enough

If the company is guiding to 84% revenue growth and a tripling of AI sales, why did the stock fall?

The answer lies in the distance between what Broadcom reported and what the most aggressive investors had already priced in. Broadcom shares had run up sharply heading into the print — the stock tested the $500 level in the days before earnings — on bets that its AI guidance would blow past even optimistic forecasts. When the Q3 AI outlook landed at $16 billion, a number that is spectacular in absolute terms, it still fell short of the most bullish buy-side models. As analysts at The Motley Fool put it, the sell-off reflected not deteriorating fundamentals but expectations that had already been stretched to their limit.

This is the defining hazard of a “priced for perfection” stock. When a company’s valuation embeds years of flawless execution, merely excellent results can read as a disappointment. The reaction is mechanical: the good news was already in the price, so anything short of a stunning upside surprise gives traders a reason to take profits. Broadcom’s after-hours dip is a textbook example — a quarter that would have sent most stocks soaring instead produced a modest pullback, because the bar had been set that high.

What It Means for the Broader AI Trade

Broadcom is not an isolated case. It is one of the most important reads on a market that has grown heavily dependent on a handful of AI-linked megacaps, and its results echo a pattern seen across the sector this year.

The dynamic showed up clearly when Nvidia posted record revenue and still saw its stock wobble on the reaction, and it reflects a market where the AI names have become both the engine of the rally and its biggest source of fragility. With the Dow, S&P 500, and Nasdaq all closing at record highs around Broadcom’s report, the index gains have been concentrated in exactly the kind of semiconductor and AI-infrastructure stocks that now trade on expectations stretched years into the future.

There is also a physical reality underneath the financial one. The custom accelerators Broadcom sells are only as useful as the data centers, power, and manufacturing capacity available to deploy them, and the industry is bumping against hard constraints on power and advanced packaging even as orders climb. Those bottlenecks are part of why guidance — not the trailing quarter — has become the number investors fixate on. The question is no longer whether AI demand is real. It is whether it can keep accelerating fast enough to justify valuations that already assume it will.

For ordinary investors, the practical takeaway is about concentration. Retirement portfolios and index funds are now heavily weighted toward a small cluster of AI-exposed giants, which means the market’s direction increasingly rides on whether companies like Broadcom can keep beating expectations that ratchet higher every quarter. The same forces that have driven record highs also make the market more sensitive to any single guidance number that lands a shade below the most optimistic case.

What Comes Next

Broadcom’s near-term path is now tied to that $16 billion AI guidance. If the company delivers — and Tan’s commentary suggests management believes the demand pipeline supports it — the current-quarter results could reset the narrative and justify the run-up that preceded this report. If AI revenue growth shows any sign of cooling, the same “priced for perfection” math that produced Wednesday’s dip will cut harder.

The broader market will be watching the read-through. Broadcom’s customers and suppliers, its peers in custom silicon, and the cloud providers building out AI capacity all take cues from its order book. A guidance number that the company itself frames as continued momentum, but that investors treated as merely good, captures the tension at the center of the AI trade heading into the summer: the fundamentals keep getting stronger, and the expectations keep getting stronger faster.

Broadcom’s next quarterly results are due in early September.