Broadcom's Earnings Call Stumble: CEO Reads Old Script, Google No Longer Betting Exclusively on Broadcom Custom Chips
0xBroomberg
Broadcom's Q2 AI semiconductor revenue hit $10.8 billion — in line with Street consensus but well short of the buy-side's $11.5 billion target. A weak Q3 guide of $16 billion versus buy-side hopes of $18.5 billion, a botched earnings call, and Google supply-chain diversification drove the stock down 12%.
Where exactly did the quarter fall short?
Q2 total revenue: $22.187 billion, just clearing the company's own $22 billion guide. AI semiconductors: $10.8 billion, up 143% year-on-year, matching Wall Street's $10.7 billion consensus.
The gap is on the buy side: bullish funds had pencilled in $11.5 billion. This means → the "beat" only held against the sell-side bar; the money that actually moves prices was disappointed.
Q3 AI semiconductor guide came in at $16 billion — a 48% sequential jump, but Street consensus sat at $17.2 billion and buy-side targets reached $18.5 billion. That is a $2.5 billion miss versus the most optimistic estimate.
What went wrong on the earnings call?
CEO Hock opened by misreading a script from fiscal 2025 Q2 — a prior-year deck. The call's overall quality fell well below past standards; the overseas research community TMTB flagged the poor performance mid-session.
In plain terms = investors don't just buy numbers, they buy management credibility. A CEO who reads the wrong script erodes confidence that forward guidance will be met.
One positive surfaced only near the call's end: Google's TPU products are seeing a significant price increase, partially offsetting the supply-chain headwind — but the timing was too late to reverse after-hours sentiment.
What does Google's supply-chain diversification mean?
Hock acknowledged that Google's AI compute demand is scaling so fast that Google will diversify its chip supply. Broadcom cannot capture all of Google's custom ASIC (application-specific integrated circuit — a chip designed to a single client's specifications) projects going forward.
This means → Broadcom's exclusive position in Google's in-house chip supply chain is loosening, potentially compressing the long-term AI revenue growth slope.
Existing orders remain large in scale. In plain terms = the near-term book is intact, but the "sole supplier" narrative no longer holds.
Can the long-term targets still be trusted?
Management reiterated a fiscal 2027 AI semiconductor revenue target above $100 billion and predicted 2028 growth would exceed 2027.
Full-year AI revenue guidance sits at roughly $56 billion, below Street consensus of $57.5–60 billion. This reflects a pattern: the long-range vision is bold, but every near-term guide undershoots expectations.
In plain terms = the "$100 billion target" is three years away. Each quarter's guide keeps disappointing the buy side — and the market will not pay up for a distant promise.
How do software and other variables look?
Infrastructure software was a rare bright spot: Q2 revenue $7.178 billion, up 9% year-on-year, with a two-year compound growth rate steady near 16%, driven by subscription renewals generating consistent cash flow.
The legacy business ex-AI grew steadily, with a two-year compound rate of just 1.3% — a stable but inelastic base.
One open variable: client Anthropic is shifting from full-system procurement to bare-die supply (buying only unpackaged chip dies, not finished assemblies). How that order-structure change affects revenue-recognition timing is a key item institutions will track in coming quarters.
Content is for reference only, not financial advice.