KEY POINTS
- Broadcom reports fiscal Q2 2026 earnings on June 3, with company guidance pointing to $22 billion in total revenue — 47% above last year and well ahead of the $20.4 billion Street consensus.
- AI semiconductor revenue is expected to hit $10.7 billion, up 140% year-over-year, driven by custom accelerator orders from Google, Meta, Anthropic, and OpenAI.
- Watch whether management reaffirms CEO Hock Tan's March call for AI chip revenue exceeding $100 billion by 2027 — that forward number is what makes or breaks the stock at 35x forward earnings.
Broadcom reports after the bell on Tuesday, and the setup is as high-stakes as any AI earnings print this quarter.
The company guided to $22 billion in total revenue for its fiscal second quarter ending May 3, a 47% increase year-over-year that blows past the analyst consensus of roughly $20.4 billion. The gap between guidance and consensus is itself a signal: Broadcom's management has consistently sandbagged estimates, and the market has learned to price in the beat. The question is whether the beat is big enough to justify a stock that has roughly tripled from its 2024 lows.
The Custom ASIC Thesis
At the center of the report is AI semiconductor revenue, expected at $10.7 billion, up 140% year-over-year. Last quarter, the segment delivered $8.4 billion, up 106%. The acceleration reflects Broadcom's unique position as the leading designer of custom AI accelerators for hyperscalers that do not want to depend entirely on Nvidia's merchant GPU platform.
Google, Meta, Anthropic, and OpenAI are all Broadcom custom silicon customers. These companies are building proprietary AI training and inference chips — Google's TPUs, Meta's MTIA — and Broadcom designs the silicon. That makes Broadcom's AI revenue a direct proxy for whether the custom ASIC market is scaling fast enough to matter next to Nvidia's merchant GPU dominance.
CEO Hock Tan put a specific number on it during the March earnings call: line of sight to AI chip revenue — chips alone, excluding networking — exceeding $100 billion in 2027. That forecast implies a compounding trajectory that would require AI semiconductor revenue to roughly triple from the current $10 billion quarterly run rate within 18 months.
VMware and the Margin Story
Beyond AI, Broadcom's software segment — anchored by the VMware acquisition — provides a margin cushion that pure-play chipmakers lack. VMware's transition to subscription licensing has been a cash flow machine, and analysts expect software to contribute roughly $7 billion in quarterly revenue with operating margins above 60%.
The combined business model gives Broadcom a diversified earnings base: hardware upside when AI spending accelerates, software stability when it does not. That dual structure is why AVGO trades at a premium to most semiconductor peers despite its lower top-line growth rate relative to Nvidia.
What Could Go Wrong
The risk sits in the guidance. If Tan walks back the $100 billion 2027 target — or hedges it more aggressively than in March — the market will read it as demand softening for custom ASICs. Hyperscaler capital expenditure plans for 2027 are still forming, and any signal that customers are shifting spending toward Nvidia's off-the-shelf Blackwell or Vera Rubin racks at the expense of custom silicon would pressure Broadcom's multiple.
Gross margin trajectory is the other flashpoint. AI semiconductor margins have been strong, but scaling production of custom chips for four or five major customers introduces execution risk that does not exist in Nvidia's one-chip-fits-all model. Analysts are watching for any margin compression as AI revenue mix shifts toward higher-volume, potentially lower-margin inference chips.
The June 3 print lands 48 hours after Jensen Huang's COMPUTEX keynote, which confirmed Vera Rubin production. That timing makes Broadcom's report a referendum on the custom-versus-merchant GPU debate. If Broadcom delivers $10.7 billion in AI revenue and reaffirms the 2027 target, it validates the thesis that the AI chip market is big enough for two architectural approaches. If it misses or hedges, the rotation risk back toward Nvidia becomes acute.
