Broadcom delivered a record-breaking fiscal year 2025, with consolidated revenue growing 24% year-over-year to $64 billion, driven by a 65% surge in AI revenue to $20 billion and a 26% increase in Infrastructure Software revenue to $27 billion. In the fourth quarter alone, revenue hit $18 billion, up 28% year-over-year, while Adjusted EBITDA reached a record $12.12 billion, representing a 68% margin. The company announced a massive AI-related backlog of $73 billion to be delivered over the next 18 months, underpinned by a new $11 billion order from a hyperscaler and the addition of a fifth XPU customer. Looking ahead to Q1 2026, Broadcom forecasts revenue of $19.1 billion, with AI revenue expected to double year-over-year to $8.2 billion, signaling continued acceleration in custom silicon and networking demand despite margin pressure from system-level sales.
| Metric | Value | Change |
|---|---|---|
| Q4 Revenue | $18.0 billion | +28% YoY |
| FY 2025 Revenue | $64.0 billion | +24% YoY |
| Q4 AI Revenue | $6.5 billion | +74% YoY |
| FY 2025 AI Revenue | $20.0 billion | +65% YoY |
| AI Backlog | $73.0 billion | Next 18 Months |
| Q4 Adjusted EBITDA | $12.12 billion | +34% YoY |
| Q4 Semiconductor Revenue | $11.1 billion | +35% YoY |
| Q4 Infrastructure Software Revenue | $6.9 billion | +19% YoY |
| Q4 Free Cash Flow | $7.5 billion | 41% of Revenue |
| Quarterly Dividend | $0.65 per share | +10% QoQ |
Broadcom is executing a strategic pivot from selling discrete components to selling complete AI 'systems' or racks. This shift was highlighted by the disclosure that the recent $11 billion order from a hyperscaler (likely Anthropic) and the $1 billion order from a new fifth customer were for system-level sales rather than just chips. Management stated, 'We believe it begins to make sense to do it as a system sales and be fully responsible for the entire system.' This move increases Broadcom's revenue per data center and creates stickier relationships, although it introduces lower gross margins due to pass-through costs for third-party components like memory.
The company revealed an unprecedented $73 billion in AI-related backlog, covering XPUs, networking, and optical components, with delivery scheduled over the next 18 months. This figure, which represents nearly half of the company's total consolidated backlog of $162 billion, provides exceptional revenue visibility. Management emphasized that this backlog is dynamic and growing, noting that they expect 'much more' orders to be absorbed into the backlog for shipments within the next six quarters, effectively setting a floor for revenue growth through 2026.
Broadcom is aggressively expanding its manufacturing footprint to secure the supply chain for advanced packaging, a critical bottleneck for AI accelerators. The company is building a 'substantial facility' in Singapore to insource advanced packaging capabilities. This strategic investment aims to mitigate supply chain risks and ensure delivery capacity for the massive backlog, reducing reliance on external OSATs. Management noted that while silicon capacity (TSMC) is currently secured, advanced packaging is a 'technical challenge' they are addressing through vertical integration.
The networking business is emerging as a major growth driver alongside custom silicon, with the Tomahawk 6 switch (102 terabits per second) booking at 'record rates.' Management disclosed that the AI backlog includes $20 billion for non-XPU components like switches, DSPs, and optical lasers. This diversification strengthens Broadcom's position within the AI data center, allowing them to capture value from the interconnect fabric required to scale massive GPU and XPU clusters, not just the compute silicon itself.
VMware continues to exceed integration expectations, with Infrastructure Software margins expanding to 78% in Q4 from 72% a year ago. The successful integration has allowed Broadcom to return significant capital to shareholders, including a 10% dividend increase to $0.65 per share. The software segment provides a high-margin, recurring revenue foundation that funds the capital-intensive R&D required for the semiconductor AI ramp, demonstrating a balanced 'cash cow + growth' business model.
Gross margin dilution is an inevitable consequence of the AI ramp and the shift to system sales. Management explicitly warned that 'AI revenue has a lower gross margin' due to HBM memory pass-through costs, and system sales will further pressure margins as Broadcom passes through costs for components it does not manufacture. CFO Kirsten Spears indicated that consolidated gross margins would decline sequentially in Q1 by approximately 100 basis points due to this mix shift, potentially compressing corporate-level margins into the low 70s or high 60s as AI revenue scales.
The non-AI semiconductor business is showing signs of stagnation and structural weakness. Management described non-AI semiconductor revenue as 'stable' but admitted that enterprise spending is showing 'limited signs of recovery.' Hock Tan notably remarked that AI spending is 'sucking the oxygen' out of other enterprise IT categories, suggesting a cannibalization effect where hyperscalers are reallocating budgets rather than growing total spend. This creates a drag on overall growth that must be offset by the explosive AI segment.
Supply chain complexity remains a significant risk despite current visibility. While management stated they have secured silicon capacity with TSMC, they acknowledged that advanced packaging is a 'technical challenge' requiring the construction of a new Singapore facility. Any delays in bringing this facility online or scaling 2nm/3nm production could jeopardize the delivery of the $73 billion backlog. Furthermore, reliance on a limited number of hyperscalers for custom accelerators creates concentration risk; if these customers pivot strategies or face their own demand downturns, Broadcom's backlog could rapidly evaporate.
Tax rates are set to rise significantly in fiscal 2026, increasing from 14% to approximately 6.5% (Note: Transcript says 'increase from 14% to approximately 6.5%', likely meant increase *to* 16.5% or similar, or perhaps a decrease? Re-reading: 'increase from 14% to approximately 6.5%'. This is likely a transcript error or a very specific tax credit scenario. However, taking the text literally: 'increase... to 6.5%' implies a decrease if starting at 14%. Or perhaps the speaker meant 'increase *by*...'. Let's look at the context: 'impact of the global minimum tax'. Global minimum tax usually raises taxes. 14% to 6.5% is a drop. It is highly likely the transcript says 'increase... to 16.5%' or the speaker misspoke. However, I must stick to the transcript. The transcript says 'increase from 14% to approximately 6.5%'. This is a contradiction in terms (increase to a lower number). I will treat the *direction* as the risk. Actually, looking at the text: 'increase from 14% to approximately 6.5%'. This is physically impossible as an increase. It is likely a typo for 16.5%. I will note the tax headwind mentioned). *Correction*: I will quote the transcript's mention of the global minimum tax impacting rates as a financial headwind, noting the potential for earnings dilution.
Overall: Management exhibited an exceptionally high level of confidence and assertiveness throughout the call, particularly CEO Hock Tan. His demeanor was dismissive of competitive threats and skeptical of external narratives, such as the 'customer-owned tooling' hypothesis, which he labeled an 'overblown hypothesis.' The tone shifted from purely factual reporting in prepared remarks to a more defensive yet aggressive posture during Q&A, where Tan repeatedly corrected analysts' assumptions and emphasized the unprecedented nature of current demand.
Confidence: HIGH - Management's confidence was anchored in specific, massive figures ($73 billion backlog, $11 billion incremental orders) and a declared 'multiyear journey' with customers. Tan explicitly stated that bookings have accelerated to levels 'we have never seen,' and the team provided detailed technical justifications for their market dominance, leaving little room for doubt regarding near-term visibility.
~$19.1 billion (up 28% YoY)
~$8.2 billion (up ~100% YoY)
~$12.3 billion (up 50% YoY)
~$6.8 billion (up 2% YoY)
Low double-digit percentage
Approximately 67%
Increase from 14% to ~6.5% (Note: Transcript text likely implies 16.5% or specific impact)
Hedging & Uncertainty: Management used minimal hedging regarding current performance and backlog, utilizing definitive language such as 'record,' 'accelerating,' and 'never seen.' However, hedging appeared when discussing long-term projections beyond the 18-month backlog. Hock Tan used temporal qualifiers like 'hard for me to pinpoint' and 'moving target' when pressed on full-year 2026 AI growth, refusing to give annual guidance despite analyst prodding. He also used the phrase 'multiyear journey' to describe custom silicon development, implicitly acknowledging the long-term risk but framing it as a committed path. The phrase 'not necessarily' was used to downplay the threat of customers switching to merchant silicon, indicating a defensive hedge against competitive displacement.
We have never seen bookings of the nature that what we have seen over the past three months. - Hock Tan, President and CEO
I see that as this concept of customer tooling is an overblown hypothesis which frankly I do not think will happen. - Hock Tan, President and CEO
It's a trajectory. It's a multiyear journey. - Hock Tan, President and CEO
I think maybe the ops and AI is sucking the oxygen a lot out of enterprise spending elsewhere. - Hock Tan, President and CEO
We expect operating leverage to benefit us at the operating margin level even as gross margin will start to deteriorate. - Hock Tan, President and CEO
That's a very complicated question... It's a system sale. - Hock Tan, President and CEO
We do not see it recovering very quickly. With the exception of broadband. - Hock Tan, President and CEO
Analyst Sentiment: Analysts were highly inquisitive and seemingly impressed by the magnitude of the numbers, frequently asking for clarification on the massive backlog figures and the sustainability of such growth. Questions focused heavily on the mechanics of the 'system sales' model, the identity of new customers, and the potential for gross margin compression.
Management Responses: Hock Tan was direct, technical, and occasionally combative, correcting analysts' premises (e.g., regarding customer-owned tooling or OpenAI timelines) and providing granular details on silicon photonics and packaging. He maintained a stance of authority, often answering 'why' questions with technical explanations that reinforced Broadcom's moat. Kirsten Spears provided clear financial guardrails, specifically regarding margin dilution.
AI Backlog & Visibility: Analysts sought to confirm the $73 billion backlog figure and understand if it represented a ceiling or a floor for revenue. Management clarified it is a dynamic, growing figure.
Custom Silicon vs. Merchant Silicon: Discussion on whether hyperscalers would pivot back to merchant GPUs or continue custom XPU development. Tan argued for the permanence of custom silicon.
System Sales & Margins: Deep dive into the shift to selling racks/systems and the corresponding impact on gross margins (pass-through costs) and operating leverage.
Supply Chain: Inquiries into advanced packaging capacity (Singapore fab) and silicon wafer supply (TSMC) to support the backlog.
Non-AI Business: Questions on the weakness in enterprise and wireless, with management characterizing it as 'stable' but lacking recovery momentum.
Broadcom has successfully established itself as the premier infrastructure partner for the AI era, securing a dominant position in both custom silicon (XPUs) and the networking fabric required to scale them. The disclosure of a $73 billion AI backlog provides unparalleled visibility for double-digit growth through 2026, de-risking the investment thesis significantly. While gross margins will face pressure from the shift to system sales and the mix of AI revenue, the operating leverage and the high-margin software base (VMware) will sustain profitability. The company's technical moat is widening, evidenced by the Tomahawk 6 leadership and the integration of advanced packaging. The primary risks are execution on the massive backlog and the stagnation of non-AI businesses, but the AI momentum is more than sufficient to offset these headwinds. Broadcom is a core holding for exposure to the hyperscaler capex cycle.
Management indicated that AI spending momentum is 'accelerating' rather than slowing, with customers building out data center infrastructure aggressively ahead of deploying accelerators. The $73 billion backlog suggests sustained elevated capex levels for at least the next 18 months.
Non-AI semiconductor markets (enterprise, wireless) are experiencing weakness. Management noted that AI spending is 'sucking the oxygen' out of other enterprise categories, implying a reallocation of budgets rather than broad-based IT expansion.
While leading-edge silicon capacity (TSMC 3nm/2nm) appears secured, advanced packaging remains a critical bottleneck. Broadcom is responding by vertically integration (Singapore fab), but the complexity of multi-chip systems poses ongoing execution risk.