Broadcom Pivots From Chip Cycle Story To AI-Driven Bet

Broadcom's business narrative has fundamentally shifted in less than a year, and the numbers tell the story. In September 2024, management led earnings commentary by framing the semiconductor cycle itself, noting that "non-AI semiconductor revenue has stabilized." A quarter later, in late 2024, management quantified the cyclical trough: non-AI semiconductor revenue hit a bottom of $17.8 billion for the full fiscal 2024. The company presented itself as a diversified operator with two reporting segments worth watching—the cyclical non-AI franchise and the execution on the VMware integration, which management called largely complete in December 2024.

By fiscal Q2 2026 (reported in early June), the composition had inverted. AI semiconductor revenue reached a record $10.8 billion, up 143% year-over-year, and accounted for 49% of total company revenue. Non-AI semiconductors came in at $4.2 billion, up a modest 6%. In absolute terms, AI revenue is now roughly 2.5 times the non-AI chip business. Management's forward guidance crystallizes the shift: fiscal Q3 2026 AI semiconductor revenue is guided to $16 billion versus non-AI at about $4.5 billion. For the full fiscal 2026, management expects AI semiconductor revenue to exceed $100 billion. The language shifted with the figures. What management now leads with are AI bookings of more than $30 billion in a single quarter (against $10.8 billion actually shipped in fiscal Q2) and a new platform with Apollo and Blackstone designed to deploy more than 20 gigawatts of compute capacity.

The non-AI business, while no longer in the lead, shows signs of cyclical recovery. Bookings topped $6 billion in fiscal Q2 2026 against the $4.2 billion actually shipped, a gap management interprets as the path to a full recovery. The fiscal Q3 guidance of about $4.5 billion represents 12% year-over-year growth, double the 6% growth just reported. Broadcom's overall financial health has strengthened: revenue grew 32% over the trailing twelve months, faster than its 29% three-year average pace; net margin of 39% is well above its 28% three-year average; operating margin of 44% sits above its 38% average.

For shareholders, the consequence is a silent shift in portfolio composition. Those who bought Broadcom as a diversified chipmaker with an AI upside now hold a concentrated bet on a handful of large AI compute buyers—a change in business identity that occurred through market forces rather than investor decision. The non-AI franchise still matters operationally, but it now receives a paragraph instead of the lead. The verdict on this shift will arrive on September 2, 2026, when fiscal Q3 2026 results are reported. If the non-AI segment clears management's guide of $4.5 billion, it will have recovered to the level it bottomed at two years prior—real, but not compounding. If it falls short, the AI story stops being one leg of the investment thesis and becomes nearly all of it, cementing Broadcom as a pure-play bet on AI infrastructure rather than the balanced operator it presented itself as in late 2024.