On June 3, Broadcom reported the largest quarter in its history and watched the stock slide after hours. The company posted $22.19 billion in total revenue for its fiscal second quarter ended May 3 — up 48% year-over-year — beat adjusted earnings estimates, guided Q3 above consensus, and still ended the session down roughly 5%.
The numbers were not the problem. AI chip revenue hit $10.8 billion, up 143% year-over-year, beating the company's own forecast. Free cash flow reached $10.26 billion for the quarter, representing 46% of total revenue. Net income increased to $9.31 billion, or $1.91 per share, up 88% from $4.97 billion, or $1.03 per share in the same quarter a year earlier.
The miss that wasn't
Wall Street wanted a raise. CEO Hock Tan did not raise the company's forecast for AI semiconductor sales in 2026, reiterating only that AI semiconductor revenue guidance for fiscal year 2027 would be "in excess of $100 billion." When a stock has climbed more than 14% in four sessions on the way into earnings, holding a forecast steady reads as a ceiling.
The stock dropped because the company narrowly missed the Wall Street total revenue consensus of $22.27 billion with $22.19 billion, and its infrastructure software unit posted $7.18 billion against a $7.32 billion expectation. That software segment, built largely around VMware, brought in $7.18 billion — missing the $7.32 billion analysts wanted, even though it grew 9% from a year ago.
Shares of the chip designer had been on a tear over the past five sessions, hitting multiple all-time highs as investors priced in solid results. "The bar was really high going into the print here," CFRA Research senior vice president Angelo Zino told Yahoo Finance as the stock was falling in after-hours.
The Q3 guide and the customers behind it
Broadcom guided Q3 fiscal 2026 revenue to approximately $29.4 billion, an 84% increase year-over-year and above the Wall Street consensus of roughly $28.53 billion. AI semiconductor revenue for Q3 is expected to reach approximately $16 billion, representing growth of more than 200% compared to the same quarter in fiscal 2025.
CEO Tan said on an earnings call that Broadcom has six core custom chip customers, including Anthropic, Google, Meta and OpenAI, which are driving the company's growth in AI revenue. In December, Tan said that Anthropic had placed an order for $10 billion in AI chips. Those relationships underpin the forward guide — Broadcom's custom accelerator model creates long-cycle engagements that are difficult to replicate or exit. The company designs custom AI accelerators, or XPUs, plus networking for specific hyperscaler clients. That model creates sticky multi-year relationships. A custom chip is co-engineered with the customer, so switching costs are high.
For small businesses building on top of the hyperscaler stack — whether through AWS, Google Cloud, or Azure — Broadcom's earnings serve as a useful cost signal. The companies paying $10 billion for custom silicon are also the ones setting the API pricing, rate limits, and inference speeds that smaller teams depend on. When Broadcom guides AI chip revenue to triple in a single quarter, the infrastructure those teams rent is what's being built. The bill hasn't arrived yet; the capex has.
