Broadcom targets $100B AI revenue
- Broadcom said on March 4, 2026, it had line of sight to more than $100 billion in annual AI chip revenue by 2027. (investors.broadcom.com) - Hock Tan said Broadcom’s first-quarter AI revenue reached $8.4 billion, up 106% year over year, and forecast $10.7 billion for fiscal second quarter. (investors.broadcom.com) - Broadcom’s next checkpoint is fiscal second-quarter 2026 results, after guiding to about $22.0 billion in revenue and 68% adjusted EBITDA. (investors.broadcom.com)
Broadcom put a concrete number on its AI ambitions in March. The chip and software company said on March 4, 2026, that it had “line of sight” to AI chip revenue in excess of $100 billion in 2027, a target Chief Executive Hock Tan tied to demand for custom accelerators and networking parts. (investors.broadcom.com) The company also said it had secured the supply chain needed to support that plan. First-quarter AI revenue reached $8.4 billion, up 106% from a year earlier, and Broadcom forecast $10.7 billion in AI semiconductor revenue for its fiscal second quarter. The number matters because Broadcom is not selling the same AI story as Nvidia. Broadcom’s role is to help large cloud and model companies design and ship custom silicon — often called ASICs or XPUs — and to pair those chips with networking and other components that move data inside AI clusters. (investors.broadcom.com) CNBC reported that Tan cited progress with Anthropic, Meta and OpenAI on the March earnings call, while saying Broadcom had secured the supply chain required to reach the 2027 target. ### Where does the $100 billion claim come from? March 4 is the date Broadcom made the claim in its first-quarter fiscal 2026 earnings release and conference call. In the release, Tan said first-quarter AI revenue was $8.4 billion and that growth was being driven by “robust demand for custom AI accelerators and AI networking.” On the call, he said Broadcom had line of sight to AI revenue from chips “in excess of $100 billion in 2027.” (investors.broadcom.com) CNBC reported that Tan also said Broadcom had secured the supply chain required to achieve that figure. CRN, citing the same earnings event, reported Tan said the company had “fully secured capacity” for relevant components from 2026 through 2028. (cnbc.com) ### What is Broadcom actually selling into AI? Broadcom’s semiconductor business includes custom AI accelerators, high-speed networking silicon and other data-center components. CNBC described the company’s role as helping other companies translate chip designs into silicon, supplying intellectual property and backend technologies before production at foundries such as Taiwan Semiconductor Manufacturing Co. December 2025 gave a fuller picture of the order book. Broadcom said its AI backlog had reached $73 billion and expected that backlog to be delivered over the next 18 months, according to multiple earnings-call transcripts surfaced by financial data sites. (investors.broadcom.com) Those orders covered not only XPUs but also switches and other AI data-center parts. (cnbc.com) ### Which customers are tied to that build-out? Anthropic, Meta, OpenAI and Google have all been named in reporting around Broadcom’s custom-chip work. CRN reported Tan discussed demand from six large customers, including Google, Meta, Anthropic and OpenAI. CNBC separately reported Tan touted progress from Anthropic, Meta and OpenAI on the March 4 call. (cnbc.com) December reporting added Google and Anthropic to the picture more explicitly. CNBC reported on Dec. 11, 2025, that Broadcom said Anthropic was the previously unnamed customer behind a $10 billion order tied to Google’s tensor processing units, or TPUs, and that Broadcom had added a fifth custom-chip customer. ### Why are custom chips getting this much attention? Custom silicon has become more attractive as the largest AI companies look for lower cost and tighter control over how models are trained and served. (fool.com) Broadcom’s pitch is that it can help those customers build application-specific chips and the networking around them, rather than relying only on general-purpose GPUs. That framing comes from Broadcom’s own description of demand for “custom AI accelerators and AI networking,” and from CNBC’s account of hyperscalers increasingly designing customized chips. (cnbc.com) Tan has also framed the competition directly. CRN reported he told investors that customers need the best chip available because they are competing with other large-language-model players and with Nvidia, which he said was continuing to improve its products each generation. (cnbc.com) ### What should investors watch next? Broadcom’s next near-term marker is fiscal second-quarter 2026. The company guided to about $22.0 billion in total revenue for the quarter, up 47% from a year earlier, and said adjusted EBITDA would be about 68% of revenue. Tan said AI semiconductor revenue should reach $10.7 billion in that period. The longer-dated milestone is 2027, when Tan said Broadcom expects AI chip revenue to exceed $100 billion. (investors.broadcom.com) Between now and then, investors will be watching whether the company converts its backlog, expands its list of custom-silicon customers and maintains the supply commitments Tan said were already in place. Those checkpoints are likely to show up first in Broadcom’s quarterly earnings releases and conference calls. (crn.com)