Broadcom's AI Chip Guidance Falls Short of the Street's Most Aggressive Expectations, Stock Drops Over 15% After Hours

Alina Collins
Published 2026-06-03About 12 min read

Broadcom posted Q2 revenue of $22.19B and AI chip revenue of $10.8B, both beating estimates, but its Q3 AI chip guidance of $16B fell short of the Street's most aggressive forecast of $17.2B — shares slid over 15% after hours, erasing roughly $250B in market cap and nearly wiping out five sessions of gains in a textbook buy-the-rumor, sell-the-news reset.

01

The quarter beat — so why the selloff?

Revenue hit $22.19B, up 48% year-over-year, edging past the $22.12B consensus; AI semiconductor revenue reached $10.8B, up 143%, slightly above the $10.7B estimate.
Adjusted EPS came in at $2.44, up 54%; free cash flow was $10.26B, roughly 46% of revenue.
This means → the backward-looking numbers were clean across the board. The trigger for the selloff was entirely in the forward guidance.
02

Where exactly did the Q3 guidance disappoint?

Q3 AI semiconductor revenue guidance: approximately $16B, up over 200% year-over-year — but $1.2B below the analyst estimate of $17.2B.
Total revenue guidance: approximately $29.4B, above the $28.2B analyst consensus but below some buy-side models; EBITDA margin guidance of 68.0% missed the 69.1% expectation.
In plain terms = the absolute growth rate is still extraordinary, but the market had already priced in the most optimistic scenario. Anything below the ceiling became a sell signal.
03

Why is this a textbook "buy the rumor, sell the news"?

In the week before earnings, Broadcom shares rallied over 14%, rebounding sharply from March lows to an all-time high — expectations were fully baked into the price.
Futurum Group CEO Daniel Newman noted that investors wanted to hear CEO Hock Tan hint at an upward revision to the $100B AI chip target for 2027. Management held to the existing guidance and did not raise it.
This reflects a structural fragility: roughly 95% of covering analysts rate Broadcom a buy, with zero sell ratings. When consensus is that uniformly bullish, any result below the most aggressive estimate triggers outsized selling.
04

Where does the custom AI chip business stand now?

The customer base has expanded from the Google TPU anchor to Meta MTIA, Anthropic, and OpenAI — Broadcom disclosed it is developing a custom chip for OpenAI, targeting a 1.3GW deployment by fiscal 2027.
This means → custom AI ASICs — chips designed exclusively for a specific client's workload — have moved from one-off projects to a platform-scale production model, reducing single-customer revenue dependency.
The product line now runs on two tracks: custom compute ASICs help hyperscalers reduce reliance on Nvidia GPUs, while networking components — switch chips, optical interconnects — scale in lockstep with AI cluster buildouts, competing against Nvidia's closed NVLink ecosystem.
05

Why is the margin under pressure?

Google TPU is the fastest-growing segment, but it carries lower margins than Broadcom's networking and software businesses, diluting overall profitability.
Apollo and Blackstone are arranging roughly $36B in debt financing for Anthropic to fund purchases of Broadcom custom chips — Broadcom backstops the largest tranche, keeping investors focused on the link between revenue recognition timing and multi-year contract backlogs.
In plain terms = AI hardware revenue is surging, but each dollar of hardware carries thinner profit than software or networking. The "revenue up, margin flat" concern is now front and center.
06

What to watch next?

Q3 delivery: whether the $16B AI semiconductor target is met — the first real test of customer delivery commitments.
Margin trajectory: as AI hardware's share of revenue keeps rising, whether operating margins can hold above 65%.
Supply-chain bottlenecks: the pace of capacity release in CoWoS advanced packaging — a process that stacks chips and memory together — HBM, and ABF substrates will directly cap the revenue growth slope into 2027.
This reflects a pivotal shift: the core variable for Broadcom's valuation has moved from "can it win orders" to "can it deliver on time without margin erosion."

Content is for reference only, not financial advice.