Photo by Laura Ockel on Unsplash
As of June 25, 2026, Broadcom (AVGO) delivered what should have been a celebration quarter: a record $10.8 billion in AI chip revenue, up 143% year-over-year. Instead, the stock fell 12-15% in after-hours trading. According to Google News, with additional coverage from CNBC, Yahoo Finance, and Bloomberg each interpreting the same results through meaningfully different lenses, the post-earnings narrative is more complex than the selloff implies.
Thesis: A $30 billion quarterly bookings figure — nearly triple the $10.8 billion actually shipped — points to supply-constrained growth, not weakening demand. The selloff reflects valuation expectations, not a crack in the underlying AI infrastructure story.
The Earnings Picture — June 25, 2026
$30 billion. That is how much in AI semiconductor orders Broadcom collected during its fiscal second quarter (ended May 3, 2026) — almost three times what it shipped. That gap between booked demand and delivered product is arguably the most underreported figure in an earnings report that generated significant market reaction.
Total Q2 revenue reached $22.19 billion, a 48% year-over-year increase. Adjusted earnings per share came in at $2.44, clearing analyst estimates of $2.32-$2.39. The AI semiconductor segment's $10.8 billion followed 106% growth in Q1 2026 — when AI revenue totaled $8.4 billion — showing continued acceleration quarter over quarter.
The "beat or miss" framing is itself contested. CNBC reported the $22.19 billion total as a miss against its Wall Street consensus of $22.27 billion. Yahoo Finance, working from a different data aggregator, called the same figure a beat against a $22.13 billion estimate. Bloomberg took a third angle: CEO Hock Tan did not raise the company's full-year AI revenue target, and Bloomberg treated that non-raise as the story's center of gravity. These are not minor editorial differences — they reflect genuinely different consensus methodologies. Investors drawing conclusions from a single outlet during earnings season are working with an incomplete picture, and any serious stock analysis of Broadcom needs to account for the full range.
What clearly disappointed markets was Q3 fiscal 2026 AI semiconductor guidance of $16 billion. That figure implies more than 200% year-over-year growth, yet still fell short of analyst expectations of $17.2 billion. At the multiples Broadcom has traded at, merely excellent is not enough.
The Evidence — Margins, Bookings, and the Hyperscaler Commitment
Chart: Broadcom AI semiconductor revenue by quarter; Q3 FY2026 reflects company guidance. Source: Broadcom earnings reports, as of June 2026.
The financial quality beneath the headline is notable. Adjusted EBITDA (earnings before interest, taxes, depreciation, and amortization — a standard measure of operating profitability that strips out non-cash charges) reached 69% of revenue, or $15.24 billion. Free cash flow — cash remaining after capital expenditures — hit $10.26 billion for the quarter, generated from $10.49 billion in operating cash flow minus $231 million in capital spending. That is a strong margin profile for any business, let alone one growing near-50% annually.
The infrastructure software segment, primarily VMware, produced $7.18 billion — a slight miss against the $7.32 billion consensus. The miss received attention disproportionate to its size, but VMware integration economics are still maturing and the segment warrants monitoring over the next two quarters.
Broadcom reiterated full-year fiscal 2026 AI semiconductor revenue guidance of $56 billion — approximately 180% growth over fiscal 2025 — and maintained its fiscal 2027 target of over $100 billion in AI chip revenue. The path from $10.8 billion in Q2 to $56 billion full-year implies significant back-half acceleration, which the $30 billion bookings backlog suggests is demand-driven rather than aspirational.
The hyperscaler commitments backing those numbers are publicly disclosed and significant in scale. Google announced an $85 billion AI infrastructure expansion in early June 2026. Anthropic disclosed plans to scale from 1 gigawatt of TPUs in 2026 to 3 gigawatts in 2027, all running on Broadcom-designed silicon. On June 24, 2026, OpenAI unveiled its first custom inference processor — codenamed "Jalapeño" — developed in collaboration with Broadcom and targeting 2027 deployment. Early testing showed the chip delivers "significantly better performance-per-watt than current state-of-the-art alternatives," per OpenAI's disclosure. Broadcom's six core custom chip customers — including Google, Meta, OpenAI, and Anthropic — have effectively replaced Apple (AAPL, approximately 15% of total revenue) as the company's primary growth narrative in the eyes of market trends watchers.
For investors conducting sector analysis of AI infrastructure: while Nvidia (NVDA) still commands approximately 70% of the AI chip market with general-purpose GPUs, custom ASICs (application-specific integrated circuits — chips designed to excel at one specific workload for a single customer) are capturing meaningful share. Broadcom's AI semiconductor segment is projected to represent 65-70% of its semiconductor revenue by fiscal 2027 — a structural shift worth researching in the context of the broader AI infrastructure buildout.
Photo by Senad Palic on Unsplash
The Bear Case Deserves Better Than a Paragraph
CEO Hock Tan's earnings call disclosures handed bears two genuinely meaningful concerns — not talking points.
First: Tan acknowledged that Google would "likely draw on multiple chip suppliers" for AI infrastructure needs. Google is Broadcom's largest AI chip customer. Any diversification toward Nvidia, AMD (AMD), or Google's own in-house Tensor Processing Units introduces customer concentration risk (where a small number of buyers account for the bulk of revenue) into an otherwise compelling thesis. Six core custom chip customers for a business targeting $100 billion in revenue is a thin roster, and investors are watching the Google relationship closely.
Second: Tan warned that surging AI semiconductor sales are weighing on overall gross margins. Custom ASICs carry lower margins than software, and as AI chips approach 65-70% of semiconductor revenue, blended margins may compress even as absolute profit dollars grow. Morningstar's analyst raised the fair value estimate to $650 per share from $550, calling the selloff "an overreaction to conservative guidance" — but conservative guidance and margin compression are not mutually exclusive outcomes. The infrastructure software segment's slight miss against consensus is a secondary but related data point worth watching.
My read: the $30 billion bookings figure is the strongest single counterargument to the bear case. Backlogs of that magnitude do not appear in weakening demand environments. But investors who priced in $17.2 billion of Q3 AI revenue were right to reset — at Broadcom's valuation level, the margin for guidance disappointment is thin, and management has now established a precedent for conservative guidance that the market will discount accordingly next quarter.
Watchlist — Metrics and Dates to Track
- Q3 FY2026 AI semiconductor revenue (report expected September 2026): The $16 billion guide is the single most important near-term data point for investment research purposes. A beat would likely reverse the post-earnings selloff quickly; a miss would confirm the bear case on demand trajectory.
- Google supplier diversification signals: Any public indication of a shift toward AMD, Nvidia, or in-house TPUs for AI workloads would directly affect Broadcom's customer concentration profile and revenue visibility into fiscal 2027.
- OpenAI "Jalapeño" production timeline: The June 24, 2026 announcement targets 2027 deployment. Volume production confirmation would add a relatively new and sizable revenue stream that strengthens the case for the $100B+ FY2027 target.
- Gross margin disclosures: Investors are watching whether the 69% EBITDA margin level holds as AI chip revenue approaches its projected 65-70% share of semiconductor revenue. Compression here would be the clearest sign that the revenue mix shift is carrying a real profitability cost.
- VMware ARR (annual recurring revenue — total software revenue contracted on a yearly basis): Closing the gap against the $7.32 billion consensus would improve earnings quality and reduce the company's dependence on the semiconductor demand cycle.
Frequently Asked Questions
Why did Broadcom stock drop after earnings even though AI revenue doubled?
As of June 25, 2026, Broadcom's Q3 fiscal 2026 AI chip guidance of $16 billion came in below analyst expectations of $17.2 billion. When a stock trades at a premium valuation — pricing in above-consensus growth — even triple-digit year-over-year gains can trigger selling if the forward outlook falls short of what the market had already priced in. This is sometimes called a "sell the guidance" reaction: the business itself did not deteriorate, but the growth implied by guidance was lower than the expectations embedded in the share price.
What does Broadcom do, and how does its AI chip business compare to Nvidia?
Broadcom designs custom ASICs (application-specific integrated circuits — chips optimized for specific AI workloads for individual hyperscaler customers) rather than the general-purpose GPUs Nvidia sells. Custom chips can deliver superior performance-per-watt for predictable, high-volume AI tasks, but require deep design partnerships and long-term volume commitments from customers. As of June 25, 2026, data suggests Nvidia still commands approximately 70% of the AI chip market. Broadcom's Q2 2026 AI revenue of $10.8 billion and $30 billion in quarterly bookings signal that the custom ASIC approach is gaining ground — investors are watching whether that share gain accelerates as hyperscalers mature their AI infrastructure programs.
What is Broadcom's AI revenue forecast for fiscal 2027, and is it credible?
Broadcom has maintained a fiscal 2027 target of over $100 billion in AI semiconductor revenue, reiterated during the June 2026 earnings call. The company's full-year fiscal 2026 guidance stands at $56 billion — approximately 180% growth over fiscal 2025. Reaching $100 billion in fiscal 2027 would require roughly doubling from that base. As of June 25, 2026, the $30 billion quarterly bookings figure suggests the demand pipeline exists to support that trajectory. Key risks investors are watching include customer concentration, gross margin compression as AI chips become a larger revenue share, and whether the OpenAI "Jalapeño" processor ramp — targeting 2027 deployment — delivers the incremental volume the $100B target implies.
Disclaimer: This article is for educational and informational purposes only. It does not constitute financial advice, a recommendation, or an endorsement of any security. Always do your own research and consult a licensed financial advisor before making investment decisions. Research based on publicly available sources current as of June 25, 2026.