Broadcom Inc. delivered exactly what Wall Street wanted on paper: fourth-quarter earnings that crushed expectations, revenue that beat estimates, and a forecast projecting AI chip sales would double in the current quarter. Yet the stock plunged more than 11% on Friday, dragging the Nasdaq lower and erasing tens of billions in market value.
The disconnect between Broadcom's strong results and savage market reaction reveals a growing tension in the AI investment thesis: can companies actually profit from the artificial intelligence boom, or are they merely spending their way to revenue growth while sacrificing margins?
The Numbers Behind the Selloff
By most conventional measures, Broadcom's quarter was exceptional. The semiconductor giant reported:
- Revenue: $14.05 billion, beating analyst expectations
- Earnings per share: Above consensus estimates
- AI revenue: Up 74% year-over-year
- Forward guidance: AI chip sales expected to double
So why the panic? Investors focused on what management didn't say as much as what they did. Gross margins, while still healthy, showed signs of compression. The cost of scaling AI infrastructure is proving more expensive than bulls had hoped.
The Margin Problem
Building the infrastructure for AI is capital-intensive in ways that traditional chip manufacturing wasn't. Custom AI accelerators require massive R&D investments, specialized manufacturing processes, and ongoing support for hyperscaler customers who demand increasingly sophisticated solutions.
For Broadcom, this means spending more to capture AI revenue. While revenue growth looks spectacular, the incremental profitability on each AI dollar is lower than the company's legacy businesses. Some analysts now worry that AI is a "good revenue, mediocre margin" story for many chip companies.
The Ripple Effect Across Tech
Broadcom's selloff didn't occur in isolation. The stock's decline weighed heavily on the broader semiconductor sector and added fuel to concerns that had been building since Oracle's disappointing report earlier in the week. Other AI-exposed names suffered:
- Nvidia: Down 3.3%
- AMD: Down 4.8%
- Micron: Down 6.7%
- Palantir: Down 2.1%
The Philadelphia Semiconductor Index fell roughly 2% on the day, extending losses from earlier in the week.
What Bulls Are Missing
Critics of the AI trade have a point: the infrastructure buildout requires enormous upfront investment, and the payoff timeline keeps extending. Companies are racing to build data centers, purchase chips, and develop models, but monetization remains concentrated among a handful of players.
The concern isn't that AI won't be transformative—most investors accept that it will be. The concern is that the companies building AI infrastructure may not capture proportional returns. They're spending billions to enable value that may ultimately accrue to the AI model developers and end users rather than the pick-and-shovel providers.
What Bears Are Missing
That said, declaring the AI trade dead after every pullback has been a losing strategy. Broadcom's AI business is genuinely booming, with custom accelerators for major hyperscalers representing a durable competitive advantage. The company's diversified portfolio—spanning networking, storage, and software—provides stability that pure-play AI stocks lack.
Moreover, the margin concerns may prove temporary. As AI chip production scales and Broadcom moves down the learning curve, manufacturing efficiencies should improve. The company has a long history of disciplined capital allocation and returning value to shareholders.
The Bigger Picture
Broadcom's selloff represents a maturation of the AI trade rather than its death. Investors are no longer willing to pay any price for AI exposure—they want to see profitable growth, not just growth. This increased scrutiny is healthy for the market, even if it produces short-term volatility.
For long-term investors, the question isn't whether Broadcom will benefit from AI—it clearly will—but whether the current valuation adequately reflects the margin dynamics of the business. With the stock now trading at a more reasonable multiple after the selloff, the risk-reward may actually be improving.
The AI infrastructure buildout is far from over. But the market is sending a clear message: execution matters, margins matter, and not every AI revenue dollar is created equal.
