By Anhata Rooprai and Zaheer Kachwala
June 3 (Reuters) – Broadcom missed Wall Street expectations for second-quarter revenue on Wednesday, as increased competition in the custom semiconductor market hurt gains from its AI chips, sending its shares down more than 11% in extended trading.
Second-quarter revenue came in at $22.19 billion, missing estimates of $22.27 billion, as Broadcom races with Nvidia whose dominant graphics processing units remain the industry standard for AI workloads.
Rivals such as Marvell Technology are also making inroads with key hyperscale customers. At the end of May, Marvell said its custom chip business would surpass $10 billion in revenue in 2029, and forecast second-quarter revenue above estimates.
The boom in inference – the process by which models respond to user queries – has made custom chips crucial to the industry, driving more orders for advanced processors and intensifying competition.
Broadcom’s ability to meet AI demand has also been tested by a strained supply chain. Company executives had in March flagged that capacity at key manufacturing partner TSMC was a “bottleneck” and that lead times for other components like printed circuit boards had stretched significantly.
“Today’s miss on revenue and subsequent post-market pull back (in shares) shows the market demands perfection for this chip rally to keep running,” said Ryan Lee, senior vice president of product and strategy at Direxion. Still, Broadcom has been one of the biggest beneficiaries of the AI race, developing custom chips for hyperscalers such as Meta and Alphabet’s Google.
Broadcom forecast third-quarter revenue of about $29.4 billion, compared with analysts’ average estimate of $28.54 billion, according to data compiled by LSEG.
“In Q3 we expect semiconductor revenue from AI to grow over 200 percent year-over-year to $16.0 billion,” CEO Hock Tan said.
That figure, however, is below estimates of $16.36 billion, according to analysts polled by Visible Alpha.
“Q2 semiconductor revenue from AI of $10.8 billion grew 143% year-over-year, above our forecast, driven by increasing demand for custom AI accelerators and AI networking,” Tan said. As the AI industry evolves rapidly, machine learning capabilities and requirements vary greatly from company to company, resulting in large cloud companies building their own processors to slash costs and personalize workloads.
(Reporting by Zaheer Kachwala and Anhata Rooprai in Bengaluru; Editing by Arun Koyyur)
By Anhata Rooprai and Zaheer Kachwala | Reuters | © Copyright Thomson Reuters 2026.
