Broadcom investors had grown used to the pattern: the company’s AI chip business prints numbers, and the stock goes up. This week the pattern broke. The company’s shares fell sharply after its fiscal-second-quarter results, erasing roughly $300 billion in market value in a single session, according to CNBC and Reuters.
The disappointment was not in the chips. Broadcom’s custom AI accelerators, built for customers including Google, and its networking silicon continue to sell as fast as the company can make them, and executives left the full-year guidance for that business unchanged. The problem was the other half of the company: software, where sales came in below expectations.
The software miss traces to VMware, the virtualization company Broadcom acquired for $69 billion in 2023. The integration has been painful from the start. Broadcom raised prices, bundled products and cut thousands of jobs, angering customers and partners, and while the strategy has improved margins, it has not produced the sales growth the company promised. This quarter, the gap showed up in the numbers.
The VMware integration has been a test of Broadcom’s playbook. The company has a record of buying mature technology businesses, cutting costs and milking them for cash, a strategy that has worked in semiconductors but faces different dynamics in software, where customers can leave. VMware’s installed base, once assumed to be captive, has proven willing to migrate to alternatives, and competitors have advertised exit programs to the company’s customers.
Wall Street’s reaction was unforgiving. Analysts had modeled steady progress in software and a possible increase in the AI guidance; they got neither. The stock’s decline reflected not just the quarter but the realization that the software story, which was supposed to make Broadcom a more stable compounder, remains a work in progress.
The AI guidance itself became a source of frustration. Investors have come to expect upgrades: every quarter for two years, Broadcom and its peers have raised their AI forecasts, and the market has come to price in the next raise. Leaving the number unchanged, even at a level that would have delighted investors a year ago, read as a disappointment.
The valuation math added to the damage. Broadcom’s stock had run far ahead of its earnings growth, with the AI narrative supporting a premium multiple. A quarter that disappointed on the non-AI side gave investors a reason to reset that multiple, and the decline reflected the reset rather than a change in the company’s core outlook, portfolio managers said.
The selloff also rippled through the sector. Other chipmakers with large AI exposure traded down in sympathy, and analysts said the move signaled a shift in how investors value the group: less tolerance for mixed messages, more demand for clean execution. The stocks that held up best were the pure-play AI names with no software division to disappoint.
The episode is the sharpest lesson the semiconductor sector has delivered this year. It shows that a strong AI story is no longer enough: companies that carry unrelated businesses must make those businesses work, or the market will punish the whole. Broadcom’s software unit was supposed to be a cushion; this quarter it was the weight.
Broadcom’s own explanation was that the software miss reflected timing, not demand. Executives pointed to contracts that slipped across the quarter boundary and said the pipeline remains healthy. The market’s response suggested it has heard variations of that explanation before, and investors are watching whether the next quarter shows actual acceleration.
The company still has the strongest position in custom AI silicon, where its design relationships with the largest cloud providers are not easily replicated. But the software business is now a visible drag on the stock, and management faces a choice: fix the VMware integration, or find a way to separate the two halves of the company.
Broadcom’s executives have signaled they see software and silicon as complementary, with AI workloads running on infrastructure built from both. But the market is pricing them separately now, and the gap between the two stories is the swing factor for the stock. The company’s next earnings call will be watched for any sign that management plans to address the software business directly, including the possibility of a divestiture.
The quarter’s numbers, taken alone, were not bad. Revenue grew, margins held, and the AI business kept compounding. The problem was the market’s expectation of more, a mismatch that has punished several technology stocks this year when guidance failed to rise.
For now, the market has voted. Broadcom’s AI story is intact, but its price is no longer being set by the AI story alone. The software numbers will decide whether the stock recovers or stays stuck, and the next quarterly report will arrive with more attention than any in the company’s recent history.
