Broadcom CEO Hock Tan.
lucas jackson reuter
Broadcom CEO Hock Tan said the artificial intelligence boom is gaining so much momentum that he predicts next year’s AI chip revenue will be “well over $100 billion.”
After the semiconductor maker reported better-than-expected first-quarter results and issued a strong outlook for the current fiscal year, Tan said on his company’s earnings call that demand is rising from large customers who increasingly need Broadcom’s help with custom silicon designs.
Regarding the sales target for 2027, Tan said, “We have also secured the necessary supply chain to achieve this.”
AI revenue more than doubled year-over-year to $8.4 billion in the first quarter, with total revenue increasing 29% to $19.3 billion. The company expects AI semiconductor sales to be $10.2 billion this quarter.
Broadcom shares soared more than 5% in after-hours trading Thursday after Tan’s comments.
Chip companies like Broadcom have faced a number of headwinds in recent months, including a lack of high-bandwidth memory needed for custom accelerators and capacity constraints in cutting-edge chip manufacturing and packaging.
Broadcom helps customers convert their chip designs to silicon and provides back-end support before the processors are manufactured at giant manufacturing plants such as Taiwan Semiconductor Manufacturing Company.
The role is one that will drive Broadcom’s growth as more tech giants design in-house accelerators for AI. Tan said custom AI implementation is entering the “next phase” and is expected to accelerate further as the company helps six major customers with their chip designs. The most likely among them are Google, Meta, Anthropic, and OpenAI, with Fujitsu and ByteDance likely to be the last two.
Google first entered the internal chip game in 2015 with a tensor processing unit designed in collaboration with Broadcom. Google has been providing its chips to cloud customers since 2018, and its current major customers include: apple And humanity. Broadcom expects demand for Google’s next-generation chips to increase further in 2027.
Meta is also reportedly in talks to use Google’s TPUs, and Broadcom is helping the social media company develop its own MTIA accelerator. Although analysts have questioned the future of Meta’s custom silicon program, “MTIA’s roadmap remains intact,” Tan said on the earnings call.
During a question-and-answer session on the conference call, Bernstein Research analyst Stacey Rasgon pressed Tan about the specific source of the projected $100 billion in AI chip revenue. He counted 3 gigawatts for Anthropic, 3 gigawatts for Google, at least 2 gigawatts for Meta, 1 gigawatt for OpenAI, and more. Tan said the amount per gigawatt “varies and sometimes changes quite dramatically,” but his estimates are “not far off.”
Tan said AI revenue growth will come from “just chips,” but Broadcom makes more than just AI accelerators. Creative Strategies’ Ben Bajarin said this includes digital signal processors, data processing units and networking switches.
It’s “everything in that bucket,” Bajarin said.
—CNBC’s Jordan Novet contributed to this report.
WATCH: Broadcom CEO talks about Ai’s earnings
