Broadcom Raises AI Revenue Guidance to $58 Billion as Custom Chip Push Accelerates

Broadcom raised its fiscal 2026 AI revenue guidance to $58 billion from a prior $56 billion, a 186% increase year-over-year, driven by surging demand for custom chips from hyperscale cloud customers. The revised forecast, announced by Chief Executive Hock Tan, signals that the company’s strategy of supplying bespoke silicon for the largest AI builders is accelerating faster than earlier expectations.

The core driver of the upgrade is Broadcom’s progress with next-generation tensor processing units. Production shipments of the TPU version 8i for Google began during the most recent quarter. Tan described the new Google TPU as performance-equivalent or superior to Nvidia’s forthcoming Vera Rubin architecture. Broadcom is also shipping the 8i ahead of MediaTek’s competing V8t version, a timing advantage that underscores Broadcom’s design and integration capabilities. Beyond Google, the company expects to accelerate shipments of its Ironwood chip to Anthropic in the fourth quarter of fiscal 2025, suggesting that the customer base for Broadcom’s AI silicon is broadening beyond its initial anchor tenant.

The guidance increase underscores a structural shift in the AI chip market. While Nvidia remains the dominant supplier of general-purpose AI accelerators, hyperscale operators such as Alphabet, Meta, and Amazon are increasingly turning to custom silicon to optimize performance and cost for their specific workloads. Broadcom has positioned itself as the primary merchant designer and supplier for these bespoke chips, leveraging its intellectual property in networking, high-bandwidth memory interfaces, and chiplet integration. The result is a rapidly growing revenue stream that is less directly tied to Nvidia’s product cycles and more tied to the expansion plans of the largest cloud builders.

For investors, the implications are significant. Broadcom’s AI revenue is now running at an annualized run rate that places it at the vanguard of the AI infrastructure build-out alongside Nvidia. The 186% year-over-year growth rate, even if partially driven by base effects from a prior low base, indicates that hyperscale capex commitments are translating into real procurement. The risk is that this revenue concentration creates dependency on a small number of very large customers. If any of those customers slow their rollout or shift internal designs, Broadcom’s growth trajectory could be interrupted.

The broader market takeaway is that the AI chip ecosystem is diversifying, but the concentration of demand among a handful of deep-pocketed buyers remains. Broadcom’s raised guidance validates the thesis that custom chips will coexist with merchant silicon for the foreseeable future. The challenge for the company will be maintaining its engineering lead and production schedules as it scales to meet these commitments, while managing the inevitable margin pressure that comes with high-volume custom designs.