Earnings Recap — Q3 FY2026
CY Q3 2026 · Reported September 2, 2026 · Beat 7 of last 7 quarters
Broadcom Inc. reported Q3 FY2026 revenue of $29.59B, a beat of 1.2% against consensus, and EPS of $3.32, a beat of 3.1%.
The Buildout is a website built on agentic AI tools, currently in beta, and could have factual errors.
Broadcom's results and multi-year outlook reinforce the accelerating AI infrastructure buildout, with custom accelerators (XPUs) becoming the preferred compute for frontier AI labs. The company's ability to secure supply and finance deployments through the XPV platform indicates that the bottleneck is shifting from chip supply to land, power, and shell. This signals sustained, multi-year demand for advanced packaging, HBM, substrates, and optical networking components.
Broadcom delivered record results in Q3 FY2026 with revenue of $29.6 billion, up 86% YoY, driven by AI semiconductor revenue of $16.7 billion, which more than tripled YoY. The company shipped Ironwood TPU v7 in high volume to Anthropic and Google, began production shipments of TPU v8i to Google, and shipped OpenAI's first-generation Jalapeno accelerator. AI networking revenue grew over 2.5x YoY. Non-AI semiconductor revenue was $4.2 billion (up 5% YoY) and infrastructure software revenue was $8.8 billion (up 29% YoY). Operating income reached a record $20.1 billion with operating margin of 67.9%, and free cash flow was a record $13.7 billion.
Management raised FY2026 AI revenue guidance to $58 billion (from $56 billion) and provided multi-year AI semiconductor revenue targets of approximately $115 billion in FY2027 and $230 billion in FY2028, both representing roughly 2x growth. They emphasized that demand exceeds these targets and that supply is secured, but cautioned that actual deployments depend on supply chain constraints. Q4 FY2026 guidance calls for consolidated revenue of $34.8 billion (up 93% YoY), AI revenue of $21.7 billion (up 236% YoY), and operating margin of approximately 66%. Gross margin is expected to decline to ~73% due to XPU mix, but operating leverage is expected to sustain operating margins. Management also highlighted the AI XPV platform with Apollo and Blackstone to finance over 20 gigawatts of compute for OpenAI and Anthropic by 2028.
“Q3 demand was simply hot, and we're just getting started.”
on Demand strength
“In 2027, we have secured the supply to again double AI revenue to approximately $115 billion. Our demand actually exceeds this outlook, and we will work to improve supply.”
on AI revenue outlook
“The lesson here is when you co-develop a chip that is optimized for your particular LLM workloads, you will outperform any GPU.”
on XPU advantage
You talked about the business doubling next year, and you said demand could be higher than that. Can you talk about the supply around that? And what are the supply bottlenecks?
Hock Tan said that demand could support shipping more, but they are being conservative and have secured supply for the $115 billion outlook. He noted that they consider whether customers can deploy chips on a timely basis, and if circumstances change, they could uplift the number.
Can you verify the gigawatt counts for '27 and '28 and the revenue per gigawatt?
Hock Tan confirmed that the gigawatt counts are for specific customers and that not all gigawatts will necessarily be deployed within the fiscal years. He said the $115 billion and $230 billion targets represent what they expect to ship with high confidence, and that XPUs cost less than half of GPUs, which explains the lower revenue per gigawatt.
What gives you certainty that OpenAI and Anthropic will take Broadcom silicon in '27/'28 when they depend on cloud providers that may have different views?
Hock Tan explained that Broadcom is enabling these labs through financing vehicles like XPV, and that they are becoming hyperscalers themselves, running their own data centers. He said it makes economic sense for Broadcom to invest in them because each gigawatt of compute can generate $30 billion of ARR.