There’s a version of this story where Broadcom is boring. An old-guard chip company, diversified into software, chugging along. That version ended sometime in 2024 and hasn’t come back.
What happened on June 3rd matters. A lot.
Broadcom reported Q2 fiscal 2026 earnings with AI semiconductor revenue of $10.8 billion — up 143% year-over-year, ahead of its own forecast, and now representing roughly half of total company sales. Consolidated revenue came in at a record $22.2 billion, up 48% from the same quarter a year ago. Adjusted EBITDA hit $15.2 billion, representing 69% of revenue. These are not the numbers of a company riding a trend. These are the numbers of a company defining one.
The stock dipped after hours. Here’s the thing — the selloff had almost nothing to do with AI. A modest miss in the infrastructure software segment spooked traders who had priced in a perfect print. Management didn’t flinch. They raised Q3 guidance to $29.4 billion in revenue, representing 84% year-over-year growth, and guided AI semiconductor revenue to $16 billion next quarter — over 200% annual growth.
Slight tangent, but it matters: the dip-after-beats pattern in mega-cap tech has become almost ritualistic in 2026. Broadcom is at least the third name this earnings cycle to post triple-digit AI growth and still sell off on a single soft line item. That’s a market running hot enough to punish perfection — which historically has been a signal worth paying attention to.
The numbers looking forward are what should hold your attention. Management reiterated full-year AI semiconductor revenue guidance of $56 billion for fiscal 2026. For 2027, CEO Hock Tan stated the company expects AI semiconductor revenues to exceed $100 billion. The pipeline is not speculative — it’s tied to custom AI accelerator demand from hyperscalers building out next-generation data center infrastructure at a scale the industry has never seen.
- Q2 AI revenue: $10.8B, +143% YoY
- Q3 AI revenue guidance: $16.0B, +200% YoY
- FY2026 AI target: $56 billion
- FY2027 AI ambition: $100+ billion
The custom silicon angle is what separates Broadcom from the pure-GPU narrative. The company designs bespoke AI accelerators — called XPUs — that major cloud platforms increasingly prefer for specific workloads because of efficiency advantages over general-purpose GPUs. That’s a sticky moat, not a commodity trade.
Risks are real. Gross margin compression is showing up as AI semiconductor mix rises relative to higher-margin software. Tan acknowledged on the call that the mix shift is diluting overall margins even as the AI engine accelerates. There’s also customer concentration — Google was cited as drawing on multiple chip suppliers, a reminder that no single hyperscaler relationship is permanent.
But zoom out. The S&P 500’s bull market in 2026 is being driven by a narrow group of AI-connected earners, and Broadcom may be the most structurally underappreciated name in that cohort. Not because it’s cheap — it isn’t — but because the trajectory of its AI business is compounding faster than the market seems willing to price in after every earnings report.
Worth a closer look before Q3 results arrive.
This article is for informational purposes only and does not constitute investment advice. Past performance is not indicative of future results. Investing involves risk, including possible loss of principal.
