Broadcom

Broadcom in the enterprise

As of November 5, 2026. Every figure links to its source; company-reported numbers are labeled as such.

Broadcom makes money from AI mainly by co-designing custom AI chips (which it calls XPUs) for a small number of very large buyers, and by selling the networking chips that connect thousands of AI processors together. That business is exploding: AI chip revenue was $16.7 billion in the quarter ended August 2, 2026, up 221% from a year earlier, and Broadcom guided to $21.7 billion for the following quarter. CEO Hock Tan says AI chip revenue should roughly double to about $115 billion in fiscal 2027 and double again in fiscal 2028, and that demand exceeds what Broadcom can currently supply (company forecasts).

The customer list is short but enormous. Broadcom builds Google's TPU chips under a long-term agreement and now ships the next-generation TPU v8i; Anthropic, which buys Google TPU capacity through Broadcom, is on track to become its largest chip customer in 2027; OpenAI has begun receiving its first Broadcom-built accelerator under a 10-gigawatt program; and Meta has extended its custom-chip partnership with an initial 1-gigawatt commitment. That concentration is the main risk: a handful of AI labs and tech giants account for most of the growth, and Broadcom's own filing notes that Anthropic's purchases depend on Anthropic's continued commercial success.

Broadcom's enterprise software arm, built on VMware, is the part ordinary corporations buy directly. It brought in $8.8 billion in the August quarter, up 29%, and Broadcom now bundles private AI tools (for running AI models inside a company's own data center) into its VMware Cloud Foundation subscription at no extra cost. Broadcom says nine of the ten largest Fortune 500 companies run on that platform, but Broadcom expects VMware revenue to dip slightly to $8.7 billion next quarter, and software's share of Broadcom's revenue fell from 43% to 30% in a year as chips surged.