Broadcom is putting a sharper point on one of the biggest forecasts in the AI hardware race: more than $100 billion in revenue by 2027, driven specifically by chips.
Speaking to investors on March 4 during the company’s fiscal first-quarter 2026 earnings call, Broadcom clarified that its projection of “significantly in excess of US$100 billion” is not a catch-all number that includes full systems, services, or software. Instead, the company said the target refers to silicon content alone. In practical terms, Broadcom is talking about the core building blocks hyperscalers need to expand AI infrastructure at massive scale, including custom XPUs, switch chips, and DSPs.
CEO Hock E. Tan emphasized that the company believes it has real visibility into this opportunity, describing a “line of sight” to the chip-only figure. Just as importantly, he framed the outlook as something built on ongoing compute deployments by individual customers rather than a series of isolated, one-time deals. That focus suggests Broadcom is positioning its AI business around repeatable, expanding rollouts as cloud and data center operators continue to add AI capacity year after year.
The takeaway for anyone tracking AI chip revenue forecasts is that Broadcom’s $100 billion-plus ambition is being presented as a pure semiconductor story. By narrowing the definition to silicon, the company is signaling confidence that demand for hyperscaler AI chips and networking silicon can grow to truly enormous levels as AI compute ramps across multiple large customers through 2027.
