The number that stopped people cold was not the revenue beat. It was the guidance.

Broadcom posted record Q1 revenue driven by AI semiconductor strength, with Q1 AI revenue of $8.4 billion growing 106% year-over-year, above forecast, driven by demand for custom AI accelerators and networking. CEO Hock Tan then stated the company expects AI semiconductor revenue to reach $10.7 billion in Q2. For context, the total Q2 revenue guide came in at $22 billion, a 47% year-over-year increase, well above analyst consensus of around $20.4 to $20.5 billion. That gap between internal guidance and Street estimates is not a small thing. It signals management has real demand visibility, not just optimism. The 106% year-over-year growth rate already exceeded Broadcom’s own internal forecast, Tan noted. When a company beats its own numbers and then guides higher, you pay attention.

Q2 results drop June 3 after market close.

What matters is the customer stack underneath all of this. For Anthropic, Broadcom is delivering 1 gigawatt of TPU compute in 2026 with demand expected to surge in excess of 3 gigawatts in 2027, a 200%-plus year-over-year ramp from a single account. For Google, the seventh-generation Ironwood TPU is driving the 2026 trajectory, with stronger demand expected from next-generation TPUs in 2027. For Meta, Broadcom expects to deliver its first gigawatt of compute in 2027, which the company has framed as a $12 billion to $15 billion revenue opportunity. OpenAI’s first-generation XPU deploys in volume in 2027 at over 1 gigawatt of compute capacity. Each new XPU customer represents 3 to 5 years of recurring silicon revenue, which is the structural reason revenue concentration is set to decline even as absolute revenue grows. That last point is the one most analysis skips over.

Slight tangent, but it connects: Meta, Broadcom, and others are joining forces to launch a $125 million semiconductor research hub at UCLA. Most coverage treats that as a footnote. It is not. It is a signal of how deep the infrastructure commitment runs between Broadcom and its largest customers. Research partnerships at that scale do not happen with vendors, they happen with long-term platform partners.

AVGO closed at $421.86 on May 27, 2026. The all-time high closing price was $439.79 on May 14, 2026. The stock has been consolidating below that level for two weeks heading into earnings. Shares are up roughly 115% over the trailing twelve months and approximately 23.74% year-to-date as of early May 2026. A $73 billion backlog with 18 months of delivery runway provides more near-term revenue certainty than almost any other semiconductor company can claim. That backlog number is what separates Broadcom’s guidance from a best-case scenario. It is already contracted.

The skepticism worth holding onto: Broadcom’s chip business bears significant customer concentration, with a small handful of large AI customers driving the bulk of revenue and future growth. If even one of those relationships shifts direction, the revenue model looks different fast. The VMware software segment, which grew just 1% last quarter, remains a quiet drag. As the perpetual-to-subscription transition completes in late 2026, analysts expect a meaningful surge in software operating income, but that transition is not complete yet. And analyst price targets ranging from $463 to $582 imply a wide range of outcomes depending on how the AI spending cycle plays out. Evercore ISI recently raised its price target to $582.

June 3 answers whether the Q2 AI guide of $10.7 billion was conservative or aggressive. That answer, plus whatever Hock Tan says about the 2027 trajectory, will matter more than the quarterly headline number itself.

Take a closer look before the number drops.

For informational purposes only. Not investment advice.