Broadcom’s 2026 Q3 Earnings Call: AI Revenue Targets Shift, Content Per Gigawatt Stability Contradicts Prior Guidance

Date of Call: Sep 2, 2026

Financials Results

• Revenue: $29.6B, up 86% YOY
• EPS: $3.32 per diluted share, up 96% YOY
• Gross Margin: 75%, down 210 basis points sequentially, up from 74% guidance
• Operating Margin: 67.9%, up 240 basis points YOY

Guidance:

• Q4 FY26 consolidated revenue expected to be $34.8B, up 93% YOY.
• Q4 AI revenue expected to be $21.7B, up 236% YOY.
• Q4 operating margin expected to be ~66%, flat YOY.
• FY27 AI revenue expected to be ~$115B, doubling from FY26.
• FY28 AI revenue expected to be ~$230B, doubling again from FY27.
• Non-AI semiconductor revenue forecast for Q4 ~$4.4B, down sequentially.
• Infrastructure software revenue forecast for Q4 ~$8.7B, stabilizing.

Business Commentary:

Revenue Growth and AI Semiconductor Expansion:

• Broadcom Inc. reported record revenue of $29.6 billion for Q3, up 86% year-on-year.
• AI semiconductor revenue more than tripled year over year to $16.7 billion, representing 56% of total revenue.
• The growth was driven by strong demand for custom AI accelerators and the expansion of AI networking, with significant contributions from key customers like Google and OpenAI.

Operating Income and Margin Improvement:

• Operating income grew by 92% year-on-year to a record $20.1 billion, with an operating margin of 68% of revenue.
• This improvement was attributed to strong operating leverage and increased revenue from high-margin AI semiconductor products.

Infrastructure Software and Non-AI Semiconductor Performance:

• Infrastructure software revenue reached $8.8 billion, up 29% year-on-year, representing 30% of total revenue.
• Non-AI semiconductor revenue was $4.2 billion, up 5% year-on-year.
• The growth in infrastructure software was driven by increased enterprise consumption of AI, while the modest increase in non-AI semiconductors was supported by infrastructure demand.

Supply Chain and Capacity Expansion:

• Broadcom is investing in expanding substrate capacity in Singapore and is increasing capacity for EML, CW, and VCSEL factories in the US and Singapore.
• These investments aim to address supply bottlenecks and support the exponential growth in demand for AI compute infrastructure, particularly for XPUs and networking solutions.

Long-term AI Revenue Outlook:

• The company expects AI revenue to double in fiscal 2027 to approximately $115 billion and again in fiscal 2028 to $230 billion.
• This outlook is based on secured supply and strong demand from key AI customers, with plans to further expand capacity and manage supply chain constraints.

Sentiment Analysis:

Overall Tone: Positive

• We delivered an exceptional quarter... Q3 demand was simply hot... We're just getting started. Our XPU shipments... up over 3.5 times year-on-year... This strong momentum continues into Q4. We are very much on target to exceed $30 in earnings per share in fiscal 2028.

Q&A:

• Question from Blaine Curtis (Jefferies): Can you talk about supply bottlenecks and what are the variables that could drive the AI revenue number higher if resolved?
  Response: Management is conservative in its outlook; demand could be higher depending on timely deployment of chips and readiness of data center sites.

• Question from Harlan Sur (J.P. Morgan): How many of your six XPU customers are using Tomahawk Scale-Out networking, and what's the update on Tomahawk Ultra adoption?
  Response: Tomahawk 6 is deployed in nearly all AI hyperscaler customers building XPUs with Broadcom. Tomahawk Ultra is starting to be deployed in scale-up applications within clusters.

• Question from Stacy: I'm verifying gigawatt counts for 2027/2028 and the revenue per gigawatt content, and how it trends with higher-generation XPUs.
  Response: The outlined gigawatts represent deployment targets, not all will ship within the two years. Demand suggests $350B in AI semiconductor shipments for FY27/28. Revenue per gigawatt is stable at ~$20-30B due to increased power per XPU.

• Question from Jack Adair (for Ben): Can you shed more light on the maximum off-balance sheet risk for the backstop agreements (XPV platform) and residual value risk?
  Response: No new announcements; future financings will be tailored to specific labs and investors, evaluated on a deal-by-deal basis.

• Question from Vivek Arya (Bank of America): What are the gross margin impacts from mix and memory cost rise, and why aren't Google,Anthropic, and OpenAI taking more given CSP funding dependencies?
  Response: Gross margin declines with higher XPU mix due to memory content. For non-AI customers, funding is self-sourced. For labs like Anthropic and OpenAI, Broadcom provides financing to enable deployment; long-term, these labs will run their own data centers.

• Question from Tom O'Malley (Barclays): How should we think about the attach rate of Tomahawk Ultra to the ASICs you're developing?
  Response: Customers building XPUs with Broadcom are adopting Tomahawk 6 and Ultra for scale-up solutions due to Ethernet-based openness.

• Question from an Analyst: To what degree is the outlook sensitized to potential land, power, and shell constraints?
  Response: Management is engaged with customers to analyze site availability and incorporates these constraints into its forecast outlook.

• Question from an Analyst: Are deployments for Anthropic and OpenAI in FY28 expected to be financed through the XPV, and what's the timeline?
  Response: Financing will be pursued on a case-by-case basis depending on the customer's situation and needs.

• Question from Jim Schneider (Goldman Sachs): Are land/power/shell the largest constraints, and what about other supply chain constraints like memory or substrates?
  Response: Land/power/shell dictates deployment timing. Other constraints like leading-edge silicon, substrates (being addressed in Singapore), and HBM memory are also multi-dimensional challenges managed collaboratively with customers.

Contradiction Point 1

AI Semiconductor Revenue Outlook and 2027 Target

Inconsistent guidance on 2027 AI semiconductor revenue scale.

Joe (Conference Call Participant) - Joe (Conference Call Participant)

2026Q3: The outlook of AI revenue doubling to ~$115B in 2027 and ~$230B in 2028 is based on secured supply and customer deployment readiness. - [Hock Tan](CEO)

What are the supply bottlenecks and variables that could drive growth higher than the projected doubling next year if resolved? - Harlan Sur (J.P. Morgan)

2026Q2: For 2026, AI revenue is expected to be ~$56 billion... The momentum is expected to continue into 2027, with AI semiconductor revenue projected to be **in excess of $100 billion**. - [Hock Tan](CEO)

Contradiction Point 2

Revenue Content Per Gigawatt Trend

Contradiction on whether content per gigawatt is stable or increasing over generations.

Conference Call Participant - Conference Call Participant

2026Q3: The dollar content per gigawatt remains stable at ~$20-$30B despite higher chip ASPs in newer generations. - [Hock Tan](CEO)

How will dollar per gigawatt improve with subsequent XPU generations? - Joshua Buchalter (TD Cowen)

2026Q2: Revenue content per gigawatt will increase over time, driven by significant price increases for more advanced XPUs... - [Hock Tan](CEO)

Contradiction Point 3

Nature and Contribution Timeline of the OpenAI Agreement

Contradiction on whether the 10-gigawatt agreement with OpenAI is a binding commitment and its expected contribution timeline.

Blaine Curtis (Jefferies) - Blaine Curtis (Jefferies)

2026Q3: The 10 gigawatts mentioned refers specifically to Anthropic’s planned deployments. - [Hock Tan](CEO)

Is the 10 gigawatts solely for Anthropic? - Christopher Rolland (Susquehanna)

20251212-2025 Q4: The 10-gigawatt agreement with OpenAI is a binding, multiyear commitment for capacity from 2027-2029... No material contribution is expected in 2026. - [Hock Tan](CEO)

Contradiction Point 4

AI Revenue Forecast Specificity and Visibility

Shift from a broad, multi-year outlook to specific annual targets creates inconsistency in forecasting methodology.

Joe (Conference Call Participant) - Joe (Conference Call Participant)

2026Q3: The outlook of AI revenue doubling to ~$115B in 2027 and ~$230B in 2028 is based on secured supply and customer deployment readiness. - [Hock Tan](CEO)

Can you discuss the supply bottlenecks and variables that could drive growth higher if resolved? - Blayne Curtis (Jefferies)

20260305-2026 Q1: The >$100 billion revenue forecast for 2027... Demand is driven by both training and, importantly, productizing LLMs for inference. - [Hock Tan](CEO)

Contradiction Point 5

Outlook for AI Revenue Growth and Supply Constraints

Contradiction between a confident, accelerating growth outlook and a cautious assessment of multi-dimensional constraints.

Joe (Conference Call Participant) - Joe (Conference Call Participant)

2026Q3: The outlook of AI revenue doubling to ~$115B in 2027 and ~$230B in 2028 is based on secured supply and customer deployment readiness. - [Hock Tan](CEO)

What are the supply bottlenecks limiting growth beyond the projected doubling next year, and what variables could drive higher demand if resolved? - Jim Schneider (Goldman Sachs)

20251212-2025 Q4: AI revenue growth is accelerating. While Q1 is a strong indicator, the full-year fiscal 2026 growth rate is not specified. The backlog continues to ramp, supporting an accelerating trend. - [Hock Tan](CEO)