Broadcom has lifted its forecast for 2026 artificial intelligence semiconductor revenue to $58 billion, a figure the company now projects for its AI chip business in the coming fiscal year.

The revised outlook puts a hard number on what has become the fastest-growing line in Broadcom's portfolio. The company's AI semiconductor revenue spans custom accelerators built for large cloud operators alongside the networking silicon — Ethernet switching and interconnect products — that ties together large-scale AI training and inference clusters.

The $58 billion forecast is a forward-looking projection rather than booked revenue, and it represents an increase over the company's earlier guidance for the period. Broadcom has not broken down the figure publicly between custom compute silicon and networking, but both categories have been the drivers behind successive upward revisions to the company's AI outlook.

The scale of the number matters for the broader supply chain. A revenue projection of this size implies sustained demand for advanced-node logic wafer capacity, high-bandwidth memory, and advanced packaging — the input side of the custom accelerator business. Broadcom designs its custom AI chips and relies on third-party foundry capacity to manufacture them, which means the forecast functions as a demand signal for leading-edge foundry and packaging suppliers even before a single chip ships.

It also sharpens the competitive picture. Broadcom sits alongside Nvidia in the small group of suppliers capturing the largest share of AI infrastructure silicon spending. Where Nvidia sells merchant GPUs, Broadcom's growth rests largely on hyperscalers commissioning their own accelerators — a model that lets cloud operators diversify their silicon supply while keeping performance tuned to their own workloads. A rising forecast for custom silicon indicates that hyperscaler in-house accelerator programs are expanding, not plateauing.

For customers, the projection signals that demand for AI compute remains strong enough to support multi-year procurement commitments across both custom and merchant silicon. For competitors, it raises the bar for anyone betting that AI chip spending would cool after the initial wave of data center buildouts.

Broadcom's guidance now positions AI semiconductors as the dominant growth engine within the company's broader portfolio, which also spans networking infrastructure, broadband, storage connectivity, wireless and industrial chips. The trajectory implied by the $58 billion figure suggests custom AI silicon will account for a substantially larger share of the company's total revenue in 2026 than in prior years.

The next test will be execution: converting the forecast into shipped silicon depends on foundry allocation, packaging capacity and the pace of hyperscaler deployment. If deployment schedules hold, the upward revisions that have characterized Broadcom's AI guidance to date are likely to continue.