After markets closed, Broadcom – the seventh-largest company in the world by market cap – posted what it described as “record revenue, record operating profit, and record free cash flow.” And then its stock fell 14%.
That reaction says more about investor psychology in the current AI market than it does about Broadcom’s business. Understanding the gap between those two things is one of the most important skills an investor can develop right now.
What Broadcom actually reported
Let’s start with what the numbers say, because they are remarkable.
AI semiconductor revenue for the quarter ended May 3 came in at $10.8 billion – up 143% year-on-year, and roughly 49% of the company’s total revenue for the period. Almost half of every dollar Broadcom earned this quarter came from AI chips.
Total revenue reached $22.19 billion, up 48% year-on-year. Non-GAAP earnings per share came in at $2.44, beating the Wall Street consensus of $2.40. Free cash flow hit a record $10.26 billion – 46% of revenue.
For the current quarter, Broadcom guided total revenue of approximately $29.4 billion – up 84% year-on-year, and well above the $28.53 billion Wall Street had been expecting. AI semiconductor revenue for the quarter is expected to reach $16 billion, representing growth of more than 200% versus the same quarter last year.
For all of fiscal 2026, AI semiconductor revenue guidance was reaffirmed at $56 billion – approximately 180% growth on fiscal 2025. And for fiscal 2027, Broadcom reiterated its target of more than $100 billion in AI semiconductor revenue, supported by multi-year contracts with its six core hyperscaler customers.
By any rational measure, this is an extraordinary business performing extraordinarily well.
So why did Broadcom fall 14%?
Three things disappointed the market, and only one of them matters.
First, the one that matters: total revenue of $22.19 billion came in just below the $22.27 billion consensus estimate. An $80 million miss on $22 billion in revenue – 0.3% below expectations. At normal valuations, this would be statistical noise. At the premium multiple Broadcom was trading on after a 40% year-to-date gain, it was enough to trigger a selloff.
Second, the infrastructure software segment, which houses VMware after Broadcom’s 2023 acquisition, generated $7.18 billion in revenue, slightly below the $7.32 billion analyst consensus. VMware has been a complicated integration, and the software segment’s slower growth relative to AI chips is a legitimate watch item.
Third, and most importantly for market sentiment: CEO Hock Tan did not raise the $100 billion AI revenue target for fiscal 2027. Tan reaffirmed it. In a market where Nvidia and other AI chipmakers have conditioned investors to expect perpetual guidance upgrades, a maintained forecast (however enormous) reads as disappointing. That dynamic is irrational, but it is real.
Why this is a concentration problem, not a Broadcom problem
The story gets bigger than one company’s earnings.
Broadcom makes up more than 3% of the entire Nasdaq 100 Index. It carries a 7.2% weighting in the VanEck Semiconductor Fund – one of the most widely held sector ETFs in the market. When a single stock falls 14% overnight, it doesn’t just affect Broadcom shareholders. It drags the index, the ETF, and every portfolio that holds either of them.
This is the consequence of the extreme market concentration that has built up around a small number of AI-related names. When things go right, the concentration amplifies gains. When they go wrong or simply when they fail to exceed impossibly high expectations, the concentrated positions amplify the pain.
You don’t have to own a single share of Broadcom to feel yesterday’s move.
The business case for looking past the noise
Strip away the market mechanics and what you have is a company with one of the most defensible positions in the entire AI ecosystem.
Broadcom does not sell off-the-shelf chips the way Nvidia does. It designs custom AI accelerators (XPUs) built specifically for each hyperscaler client’s model architecture. Google, Anthropic, OpenAI, Meta, and two others currently ramping hard are the six core customers. These relationships are not quarterly purchasing decisions.
They are multi-year, multi-gigawatt infrastructure commitments.
The switching cost is not high. It is almost prohibitive. A custom chip is co-engineered with the customer.
Switching to a different vendor means years of redesign and billions in retooling investment for the customer.
Once Broadcom is designed into a customer’s AI infrastructure, it is effectively permanent.
On top of the custom silicon, Broadcom makes the networking plumbing – Ethernet switching, serialiser-deserialiser components, PCIe interconnects, digital signal processors, co-packaged optics – that connects every component of an AI cluster together.
You cannot build a 100,000-GPU training cluster without Broadcom’s networking. You cannot link ten data centres into a coherent AI fabric without Broadcom’s networking. As Tan noted on the earnings call:
Broadcom’s content per gigawatt of AI compute will increase materially as chips become more complex and embed more functionality. The revenue per unit of AI infrastructure is going up, not down.
AI bookings in the quarter exceeded $30 billion. Supply for 2026 and 2027 is secured. The company is already working on supply commitments for 2028 and 2029.
This is not a company whose growth is slowing. It is a company whose growth failed to exceed what the most aggressive analysts had pencilled in, and in the current AI market, those two things have been treated as equivalent.
For investors with a time horizon longer than a news cycle, yesterday’s selloff looks far more like an entry point than an exit signal. And for everyone else, it’s a reminder that in a concentrated market, you are never as insulated from single-stock risk as your diversified portfolio might suggest.
Not a subscriber to Money Morning?
You can get free daily recommendations like these with Money Morning eletter. Just sign up here.
