Broadcom reported record financial results for the fourth quarter and full fiscal year 2025, driven by accelerating demand for AI-focused semiconductors and sustained growth in infrastructure software. Fourth-quarter revenue reached $18.0 billion, up 28% year over year, exceeding company guidance as hyperscale customers continued to invest in custom AI accelerators and high-capacity Ethernet switching.
AI semiconductors led growth, with revenue increasing 74% year over year in Q4. Looking ahead, Broadcom forecast first-quarter fiscal 2026 revenue of $19.1 billion, also up 28% year over year, and expects AI semiconductor revenue to double year over year to $8.2 billion, supported by custom accelerators and Ethernet-based AI networking. Infrastructure software revenue rose 19% year over year in Q4, with VMware Cloud Foundation continuing to drive bookings and backlog expansion.
Despite the strong financial performance and forward outlook, Broadcom shares fell roughly 11%, leading the market lower as investors reacted to renewed concerns around AI infrastructure spending cycles, valuation sensitivity, and margin mix implications tied to the rapid scaling of AI systems. Management emphasized that customer demand, order velocity, and backlog visibility across AI compute, networking, and optics remain strong heading into fiscal 2026.
Key Points
- Q4 revenue: $18.0B, up 28% year over year
- FY2025 revenue: $63.9B, up 24% year over year
- Q1 FY2026 revenue outlook: $19.1B, up 28% year over year
- AI semiconductor revenue growth (Q4): +74% year over year
- Quarterly dividend: increased 10% to $0.65 per share
- Broadcom shares: down ~11% amid broader AI-related market volatility
“In Q4, record revenue of $18.0 billion grew 28% year-over-year, driven primarily by AI semiconductor revenue increasing 74% year-over-year. We see the momentum continuing in Q1 and expect AI semiconductor revenue to double year-over-year to $8.2 billion, driven by custom AI accelerators and Ethernet AI switches,” said Hock Tan, President and CEO of Broadcom Inc.
Key Points from the Investor Call
- AI backlog visibility: Broadcom reported $73 billion in AI-related backlog, covering XPUs, Ethernet switches, DSPs, lasers, and PCIe switches, expected to ship over the next 18 months, with management indicating bookings continue to accelerate beyond that level.
- Custom accelerator momentum: The company now supports five XPU customers, including a newly disclosed fifth customer with a $1 billion order for delivery in late 2026. Management characterized custom accelerators as multi-year strategic programs rather than transactional replacements for GPUs.
- XPU product diversity: Each hyperscale customer is developing multiple generations and variants of XPUs, optimized separately for training, inference, reasoning, sparse workloads, and memory bandwidth, resulting in expanding silicon content per customer over time.
- AI networking demand outpacing compute: Demand for Ethernet-based AI networking is building ahead of accelerator deployment, with AI switch backlog exceeding $10 billion, led by the 102 Tbps Tomahawk 6, which management described as one of the fastest-ramping switch products in the company’s history.
- Optical and DSP pull-through: Strong AI cluster builds are driving record orders for 1.6 Tbps DSPs, lasers, and optical components, with management noting that non-XPU AI components account for roughly $20 billion of the current AI backlog.
- Shift toward system-level sales: Broadcom is increasingly delivering full AI systems or racks, bundling XPUs, networking, optics, and interconnects, rather than selling standalone chips. This approach increases revenue scale while introducing more pass-through components.
- Margin dynamics: Management expects gross margin percentages to decline as AI systems revenue grows, but emphasized that gross margin dollars and operating income dollars will continue to expand due to volume growth and operating leverage.
- Advanced packaging strategy: To address multi-chip and advanced packaging constraints, Broadcom is partially insourcing advanced packaging through a new facility in Singapore, focused on supply-chain security and delivery rather than cost reduction.
- Process node access: The company indicated it currently sees no constraints at 3-nanometer or 2-nanometer nodes at TSMC, though capacity planning remains an ongoing focus as AI demand scales.
- Silicon photonics timing: Management stated that silicon photonics will eventually become necessary for scale-up and scale-out interconnects, but indicated that copper and pluggable optics remain viable for now, with photonics adoption expected later in the roadmap.
- Non-AI semiconductor outlook: Outside AI, management described demand as stable but subdued, with broadband showing recovery while enterprise-oriented markets remain pressured as AI spending absorbs a larger share of hyperscaler budgets.
🌐 Analysis
Broadcom’s results highlight its expanding role as a full-stack supplier to hyperscale AI infrastructure, spanning custom silicon, Ethernet switching, optical interconnects, and infrastructure software. While equity markets reacted negatively amid broader AI-spending jitters, the company’s backlog growth, product breadth, and cash generation continue to support its long-term positioning as AI deployments scale into 2026.
Investor concerns appear focused on valuation sensitivity and the near-term impact of faster-growing AI systems revenue on reported gross margins. Broadcom has consistently framed this tradeoff around absolute profit growth and operating leverage, emphasizing margin dollars rather than margin percentage as AI infrastructure spending accelerates.
🌐 We’re tracking the latest developments in networking silicon. Follow our ongoing coverage at: https://convergedigest.com/category/semiconductors/
🌐 We’re launching the “Data Center Networking for AI” series on NextGenInfra.io and inviting companies building real solutions—silicon, optics, fabrics, switches, software, orchestration—to share their views on video and in our expert report. To get involved, send a note to jcarroll@convergedigest.com or info@nextgeninfra.io.