Hey there, bargain hunter.
Broadcom reported Q2 fiscal 2026 earnings on June 3rd, and the headline number was historic. AI semiconductor revenue hit $10.8 billion – up 143% year-over-year. Total revenue came in at $22.2 billion, up 48% from the same quarter a year ago. Non-GAAP EPS of $2.44 beat the street’s $2.40 estimate. Then the stock got hammered roughly 15%.
Let that sit for a second.
The reason wasn’t the AI engine – it’s still on fire. The drag was a small shortfall in the infrastructure software segment, and CEO Hock Tan’s decision not to raise the full-year AI semiconductor revenue target beyond the already-staggering $100 billion guidance he had previously put on the table. The market wanted a raise. It didn’t get one. Cue the selloff.
What the Business Actually Looks Like Right Now
Here’s the thing: Broadcom doesn’t sell off-the-shelf GPUs the way Nvidia does. It designs custom AI accelerators – called XPUs – and the networking stack that ties them together, built specifically for hyperscaler clients. Tan confirmed Broadcom now has six core custom chip customers, including Anthropic, Google, Meta, and OpenAI. That is a sticky, co-engineered relationship. Switching costs are high. The pipeline is deep.
Q3 guidance is where it gets interesting. Management guided Q3 total revenue to approximately $29.4 billion – up 84% year-over-year and above the roughly $28.5 billion Wall Street had penciled in. AI semiconductor revenue alone is guided to grow over 200% to around $16 billion next quarter. Adjusted EBITDA is expected to hold at approximately 68% of projected revenue.
Broadcom also launched a new $10 billion share repurchase program and announced it has begun shipping the industry’s first 2nm custom compute SoC on its 3.5D XDSiP platform – a meaningful technical lead that doesn’t show up in one quarter’s numbers.
Is It Cheap After the Drop?
Analysts revised fair value estimates upward from roughly $404 to $456 earlier this year on the back of AI demand data. The stock was trading near $345 before the earnings print. Post-drop, depending on where it settles, that gap between analyst consensus and current price starts to look worth exploring for patient capital.
Slight tangent – but it matters: the market tends to punish guided conservatism harder than it rewards beat-and-raise. Tan has consistently guided below and delivered above. That pattern is worth keeping in your back pocket.
- Q2 AI revenue: $10.8B (+143% YoY)
- Q2 total revenue: $22.2B (+48% YoY)
- Q3 revenue guide: ~$29.4B (above consensus)
- Q3 AI revenue guide: ~$16B (+200%+ YoY)
- FY26 AI revenue target: in excess of $100B
- New buyback: $10B authorized
- EBITDA margin guide: ~68% of revenue
What actually happened here is straightforward. A company compounding AI revenue at triple-digit rates got hit because it held its guidance steady instead of raising it. The business didn’t break. The narrative just didn’t accelerate fast enough for a market that wants upside surprises every 90 days.
Whether the selloff creates a real entry point or just a head-fake depends entirely on whether that $100 billion AI revenue target for fiscal 2026 holds – and on Q3 execution. Watch the software segment closely. That’s where the concern lives, not in the custom silicon story.
