Broadcom Inc. (AVGO) Earnings

Management highlights

### Core Business Growth & Demand - Q2 FY2026 set records for total revenue, operating margin, and adjusted EBITDA, beating prior guidance on the strength of AI semiconductor demand. - AI semiconductor bookings in Q2 reached over $30 billion, far exceeding the $10.8 billion shipped in the quarter, giving management extended visibility into future demand (now extending to 2028, up from 2027 three months prior). - Non-AI semiconductor bookings exceeded $6 billion in Q2, signaling a full cyclical recovery for that segment. ### Strategic Customer Partnerships - Multi-generational long-term agreement with Google to develop and supply multiple generations of TPUs and AI networking, with a substantial committed value. - Additional 5 gigawatt next-generation TPU-based compute agreement with Anthropic, on top of the 1 gigawatt already planned for 2026, with deployment starting in 2027. - On track for production silicon delivery to OpenAI in late 2026, with a contractual commitment to deploy 1.3 gigawatts in 2027 as part of a larger 10 gigawatt target by 2029. - Multi-generational XPU partnership with Meta, with 3 gigawatts planned for deployment by end of 2028; initial 1 gigawatt order (including XPUs and networking) has been received, with delivery starting in H2 2027. - Initial shipments to two additional core AI customers are scheduled to begin in late 2026, with $6 billion in purchase orders received to date. ### Technology Leadership - Holds at least one generation of technology leadership in AI networking, a critical component for scalable AI clusters: - Industry-leading 200G/400G SerDes for direct-attach copper within racks, supporting co-packaged copper with Ethernet and PCI Express switches - Has shipped the industry's only 100-terabit Ethernet switch (Tomahawk 6) for over a year; next-generation 200-terabit switch will tape out this quarter - De facto industry standard for co-packaged optics (CPO), with 1.6-terabit DSPs and laser components - Leading Jericho 3/Jericho 4 fabric solutions for cross-data center AI cluster scaling, powering the world's largest hyperscale deployments ### New Strategic Initiatives - Launched the AI XPU platform in partnership with Apollo, Blackstone and other leading investors, targeting 20+ gigawatts of total compute capacity to deliver low-cost, low-power AI compute for leading frontier AI labs. The first $35 billion tranche is currently being launched by Apollo. - Released VMware Cloud Foundation 9.1, which adds heterogeneous compute support for AMD, Intel, and NVIDIA GPU/CPU architectures to enable enterprise AI inferencing and traditional workloads on a common private cloud platform; on-prem deployment demand is extremely strong.

Guidance

- **Q3 FY2026 Consolidated Guidance**: Total revenue of $29.4 billion, up 84% year on year; operating margin stable at ~67% quarter over quarter; adjusted EBITDA of ~68% of revenue; consolidated gross margin expected to decline to ~74% due to mix shift to higher semiconductor (lower gross margin than software) content, which is not a structural margin decline. - **Q3 FY2026 Segment Guidance**: Semiconductor revenue of $20.5 billion, up 124% year on year; AI semiconductor revenue of $16 billion, up over 200% year on year; non-AI semiconductor revenue of ~$4.5 billion, up 12% year on year; infrastructure software revenue of ~$8.9 billion, up 31% year on year. - **Full Year FY2026 Guidance**: AI semiconductor revenue expected to reach $56 billion, up ~180% from FY2025; H2 2026 AI semiconductor revenue expected to double H1 2026 shipments. - **FY2027 Guidance**: Reiterates prior guidance that AI semiconductor revenue will exceed $100 billion, with growth on track or stronger than previously projected. 10 gigawatts of AI compute shipments are planned for FY2027, unchanged from prior guidance, with shipments back-half loaded. - **FY2028 Guidance**: Expects substantial AI semiconductor revenue growth from FY2027 levels, supported by existing customer commitments and growing demand.

Segment performance

1. Semiconductor Solutions Segment: - Total Q2 FY2026 revenue: $15 billion, growing 79% year on year, representing 68% of total consolidated revenue. - AI semiconductor sub-segment revenue: $10.8 billion, growing 143% year on year, representing 49% of total consolidated revenue. Networking accounted for ~40% of total AI semiconductor revenue in Q2. - Non-AI semiconductor revenue: $4.2 billion, growing 6% year on year. - Segment gross margin: 70%, operating margin: 62% (up 460 basis points year on year), operating expenses: $1.2 billion (8% of revenue). 2. Infrastructure Software Segment: - Total Q2 FY2026 revenue: $7.2 billion, growing 9% year on year, representing 32% of total consolidated revenue. - ARR growth sustained at 17% year on year. - Segment gross margin: 93%, operating margin: 79% (up 310 basis points year on year), operating expenses: $1 billion. Consolidated Q2 FY2026 performance: - Total revenue: $22.2 billion, up 48% year on year. - Gross margin: 77.1% (down 32 basis points year on year, driven by mix shift to semiconductors). - Operating income: $14.9 billion, up 52% year on year; operating margin: 67.3% (up 200 basis points year on year, on flat operating expenses). - Adjusted EBITDA: $15.2 billion, 69% of revenue (above guidance of 68%). - Free cash flow: $10.3 billion, 46% of revenue.

Risks & headwinds

No explicit material risks or operational failures were discussed by management during the call. Standard forward-looking statement risk disclosures referenced recent SEC filings, but no new or specific risks were raised.

Analyst Q&A

• Q: How do the stated full-year 2026 AI revenue guidance of $56 billion and projected 2x second-half over first-half growth align, given that 2x growth would result in a higher total? Is the current 18-month AI backlog now $200 billion or more? /

  A: Hock Tan explained that the $56 billion figure aligns with the 2x projection: first-half 2026 AI shipments totaled ~$19 billion, so doubling that for the second half results in a full-year total of approximately $56 billion. He confirmed that the company remains on track (or is tracking stronger) to hit its prior target of over $100 billion in AI revenue for 2027, and declined to provide quarterly updates to the 2027 target.

• Q: What is the outlook for networking's share of AI revenue, and when will co-packaged optics (CPO) revenue become meaningful? /

  A: Tan noted that 40% networking share of AI revenue in Q2 was an unusually high alignment of factors, as the company was shipping high volumes of networking to non-XPU customers while growing XPU-related networking. He expects networking's share of total AI revenue to settle closer to ~30% over time, as growing XPU shipments increase the base of total AI revenue. He did not provide a specific timeline for meaningful CPO revenue.

• Q: Why did AI bookings hit $30 billion in the quarter, far exceeding Q2 shipments? /

  A: Tan explained that large AI frontier customers are now planning far ahead to secure capacity, as building out AI compute requires aligning lead times for wafers, HBM memory, power infrastructure, and data center connectivity. Customers are placing large orders early to lock in capacity, which gives Broadcom extended visibility (now through 2028, up from 2027 just three months prior). This advance ordering reflects insatiable demand growth, not just component shortages.

• Q: How should we think about revenue content per gigawatt of AI compute over time? /

  A: Tan confirmed that revenue content per gigawatt will increase over successive product generations. Next-generation XPUs are larger multi-die designs that incorporate more features including higher amounts of HBM, fast SRAM, and embedded CPU cores, which drive higher ASPs per chip, leading to higher total content per gigawatt of deployed compute.

• Q: Is Broadcom able to secure incremental wafer and HBM supply to meet growing customer demand, and will it add additional foundries for supply optionality? /

  A: Tan stated that Broadcom has already secured sufficient supply to meet its needs for 2026 and 2027, and is currently working to secure supply for 2028 and 2029. He confirmed that the company is generally able to secure incremental supply to meet new incremental customer demand that has emerged over recent months.