Broadcom Inc. reported record revenue and forecast artificial intelligence (AI) semiconductor sales to double in the current quarter, yet shares reversed earlier gains to close down approximately 4.5% in after-hours trading Thursday. The selloff came after CEO Hock Tan's commentary on the earnings call raised concerns about margins, OpenAI revenue timing, and non-AI business weakness, overshadowing strong fiscal fourth-quarter results.
AI Generated Image
The chip designer initially saw shares surge as much as 4% after posting fiscal fourth-quarter revenue of $18 billion and guiding fiscal first-quarter revenue to $19.1 billion, both beating Wall Street estimates. The company announced AI semiconductor revenue jumped 74% year-over-year and would double to $8.2 billion in the current revenue, significantly above the consensus forecast of $6.9 billion.
However, momentum shifted during the analyst call when Tan revealed that AI revenue carries lower gross margins than non-AI revenue and that non-AI revenue would remain flat year-over-year in the current quarter. He also indicated the company's contract with OpenAI would not yield significant returns until 2027-2029, stating "we do not expect much in 2026."
The volatile reaction highlights investor concerns that even robust AI demand may not be enough to sustain Broadcom's 75% year-to-date stock gain, particularly as questions mount about margin pressure and revenue concentration risks in the AI buildout.
Record Results Driven by AI Chip Demand
Broadcom delivered its strongest quarterly performance on record, with revenue of $18 billion in the fiscal fourth quarter ended November 2, up 28% year-over-year (YoY), and earnings per share (EPS) came in at $1.95, topping analysts expected $1.87. Adjusted EBITDA, or earnings before interest, taxes, depreciation, and amortization, hit $12.2 billion, or 68% of revenue, also a company record.
AI semiconductor revenue climbed to $8.3 billion in the fourth quarter, substantially exceeding the company’s prior guidance of $6.2 billion. For the fiscal year 2025, adjusted EBITDA increased 35% YoY to a record $43 billion, while free ash flow reached $26.9 billion. The company raised its quarterly dividend 10% to 65 cents per share.
The Palo Alto-based company guided fiscal first-quarter revenue to approximately $19.1 billion, well ahead of the $18.5 billion average analysts’ estimates. The forecast include AI semiconductor revenue of $8.2 billion, representing a doubling from the prior-year period.
Customer Orders and Backlog Details Emerge
Tan disclosed that Anthropic was the previously unnamed fourth customer that signed for at least $10 billion in custom chips last quarter. The AI startup placed an additional $11 billion order in the fourth quarter for delivery in late 2026, bringing total commitments to $21 billion.
The CEO also revealed Broadcom secured a fifth custom-chip customer during the quarter with a $1 billion order expected for late 2026 delivery, without identifying the fifth client. He said the company’s total AI-related backlog exceeds $73 billion for delivery over the next 18 months, though he emphasized this figure represents a “minimum” as more orders are anticipated.
"We do expect much more as more orders come in for shipments within that next six quarters," Tan said. "So our lead time, depending on the particular product it is, can be anywhere from six months to a year." The company's backlog for AI switches alone exceeds $10 billion.
Margin and Revenue Mix Concerns Surface
Investor enthusiasm cooled as Tan acknowledged that AI product sales are compressing total margins. He told analysts that the fast-growing AI revenue generates lower gross margins compared to the company's traditional non-AI business, which is expected to decline sequentially in the current quarter.
The shift in revenue mix toward lower-margin AI products represents a departure from Broadcom's historically robust profitability. Tan also said non-AI revenue would remain flat YoY in the first quarter, highlighting weakness in other segments amid the AI buildout.
OpenAI Deal Timeline Disappoints
Tan's comments about the OpenAI partnership, announced in October, triggered particular concern. When asked about revenue timing from the agreement to provide up to 10 gigawatts of data center infrastructure between 2026 and 2029, he said the bulk of returns would come in 2027, 2028, and 2029 rather than near term.
The delayed OpenAI revenue added to broader worries about the AI startup's financial commitments. Oracle Corp. shares tumbled more than 10% Thursday after disclosing that a bulk of its future revenue commitments came from OpenAI, making Broadcom's exposure another potential risk factor for investors.
Market Reacts to Lack of Surprises
Broadcom's decline reflects how high expectations have become after the stock's 75% gain this year. Analysts noted the company provided no major surprises on the call comparable to previous quarters when unnamed customers or unexpected order sizes drove sharp rallies.
Jake Behan, Direxion's head of capital markets, said "raw AI spending isn't enough to keep traders excited" and that shares had "come so far and so fast this year that even a solid report could end up seeing profit-taking." The stock closed regular trading at $406.37 before falling to around $385 after hours.
Tan also gave a brief response to questions about "customer tooling" — the risk that large customers might develop chips internally rather than ordering from Broadcom. He called it "an overblown hypothesis," arguing that competing against merchant GPU makers like Nvidia Corp. requires resources most AI model developers prefer to allocate elsewhere. The lack of detailed pushback on this bear thesis may have contributed to selling pressure.
Explore more exclusive insights at nextfin.ai.
