Why Is Broadcom Inc. (AVGO) Down 3.8% Since Last Earnings Report?

Broadcom's fiscal Q3 2026 non-GAAP EPS of $3.32 rose 96.4% year over year and beat consensus by 3.11%, while revenue surged 85.5% to $29.59 billion, led by a 221% jump in AI semiconductor revenue to $16.7 billion. The company guided fiscal Q4 revenue to approximately $34.8 billion (+93% year over year), including $21.7 billion of AI semiconductor revenue (+236%), and raised its fiscal 2026 AI revenue outlook to $58 billion from $56 billion. Record free cash flow reached $13.67 billion, alongside $3.1 billion in dividends and $5.6 billion of long-term debt reduction, though AVGO shares have declined 3.8% since the earnings release and carry a Zacks Hold rating.

Analysis

The relevant debate is no longer whether AVGO can grow AI revenue, but whether custom silicon can sustain a valuation premium while its mix structurally dilutes gross margin. The company’s forward capacity commentary is a demand indicator rather than contracted revenue: hyperscaler deployment schedules, HBM availability, advanced packaging, and data-center power interconnects can each defer recognition by a quarter or more. Near term, the market is likely to reward evidence that the ramp converts into earnings without another material margin step-down; a miss on gross margin or timing would matter more than a modest revenue beat.

Custom accelerators create a second-order challenge for NVDA: they reduce the addressable share of hyperscaler training/inference spend, but they do not necessarily displace merchant GPUs at the system level. The clearer beneficiaries are bottleneck suppliers—MU for HBM, TSM for leading-edge wafers/advanced packaging, and optical/interconnect vendors such as COHR and LITE—provided supply additions do not arrive ahead of end demand. MRVL is a less clean read-through: its optical and custom-silicon exposure benefits from buildouts, but its smaller scale and customer concentration leave it more vulnerable if one hyperscaler shifts architecture or delays a cluster.

Contrarian view: the recent AVGO weakness is potentially an opportunity only if it reflects positioning rather than a reset in forward AI expectations. However, the market may be underpricing the 6-18 month risk that hyperscalers use internally designed silicon to bargain down accelerator pricing and that memory-rich configurations keep mix pressure elevated. Thesis is falsified by another meaningful reduction in semiconductor gross-margin outlook, a disclosed customer deployment delay, or a sequential slowdown in AI revenue despite supply availability.

Key Decisions for Investors

• Accumulate AVGO on weakness into the next earnings window, preferably via a 3-6 month call spread rather than outright shares; target a 10-15% upside on continued estimate revisions, with a 7-8% risk limit if management cuts gross-margin guidance or signals a material hyperscaler timing slip.
• Pair trade over 1-3 months: long AVGO / short MRVL in equal semiconductor-beta terms. AVGO has broader customer diversification, software cash-flow support, and balance-sheet flexibility; cover if MRVL delivers a material custom-silicon design-win disclosure or AVGO AI revenue guidance misses by more than 5%.
• Maintain a watch-list long in MU and TSM rather than chase immediately; initiate only after confirming HBM pricing/bit-growth and advanced-packaging lead times remain tight. These are higher-purity bottleneck expressions of continued custom-accelerator deployment, but are vulnerable to a synchronized hyperscaler capex pause.
• Do not treat GOOG or META as direct near-term beneficiaries without capex and monetization evidence. Their internal-silicon programs can improve long-run compute economics, but near-term incremental spending remains a cash-flow and depreciation headwind; reassess after their next capex guidance and AI-product revenue disclosures.