Key Highlights

  • Broadcom (Nasdaq:AVGO) Q2 FY2026 Earnings due after market close; Revenue consensus at $22.1 billion, up 47% year over year.
  • AI semiconductor revenue projected at $10.7 billion, up 140% year over year, accelerating from Q1's $8.4 billion.
  • Broadcom holds approximately 70% of the global ASIC market with customers including Alphabet, Meta, OpenAI, Anthropic, and Apple, six XPU customers as of Q1 FY2026.
  • Stock trades at a P/E ratio of 92.31 near its 52-week high, leaving no Margin for an execution shortfall.
  • Management's $100 billion AI chip revenue target for 2027 faces its most rigorous test on tonight's call.

A Company That Has Earned the Market's Attention

Broadcom Inc. is not a story that emerged overnight. It is the product of disciplined engineering, strategic acquisitions, and a decade-long bet that the largest technology companies in the world would eventually want chips designed specifically for them. Tonight, when it reports Q2 FY2026 financial results after the US market close, that bet faces its most consequential quarterly test yet.

The numbers expected are substantial. Wall Street's consensus sits at approximately $22.1 billion in total revenue and $10.7 billion in AI semiconductor revenue for the quarter ending May 3, 2026. Both would represent records. Both would confirm that the trajectory established in Q1, when AI revenue grew 106% year over year to $8.4 billion, is not decelerating.

The stock is down 1.29% at $475.38 in early trading on June 3, a movement more consistent with pre-earnings caution than any fundamental concern. The Options market is pricing in a post-earnings move of approximately 10.65% in either direction. For a company with a Market Capitalisation of $2.25 trillion, that range reflects the scale of what tonight's print will decide.

The Custom Silicon Model and Why It Is Difficult to Replicate

Broadcom's competitive position in AI infrastructure is built on a specific product category: application-specific integrated circuits, or ASICs. Unlike the general-purpose graphics processing units that power the majority of AI Training and inference workloads globally, ASICs are custom-designed for a single customer's architecture. They do one set of tasks with greater energy efficiency and lower total cost than a standardised chip at the same scale.

The Business model carries unusual structural advantages. Designing a custom chip for a hyperscaler requires years of joint engineering between Broadcom and the customer. The resulting product is architecturally inseparable from the buyer's infrastructure. Migrating away from it is not a procurement decision; it is a multi-year re-engineering project. That reality underpins the kind of revenue durability that standardised chip vendors cannot credibly claim.

Broadcom holds approximately 70% of the global ASIC market. It's customer base has widened, confirmed XPU partnerships now span Google, Meta, OpenAI, Anthropic, and Apple, with the Apple relationship publicly disclosed for the first time in 2026.

Advanced Micro Devices Inc. (NASDAQ:AMD), Intel Corporation (NASDAQ:INTC), and Marvell Technology Inc. (NASDAQ:MRVL) are active in the space, but none has assembled the same combination of design depth, Manufacturing partnerships, and installed customer base. The moat is real, and it has been built over time rather than acquired quickly.

Alphabet, Meta, OpenAI and Anthropic: What Their Spending Tells Us

Broadcom's earnings are not a narrow semiconductor story. They are a direct read on Capital allocation at the highest levels of the global technology industry. The scale of that capital is now definitively large. The four largest hyperscalers: Alphabet Inc. (NASDAQ:GOOGL), Meta Platforms Inc. (NASDAQ:META), Amazon (NASDAQ:AMZN), and Microsoft (NASDAQ:MSFT), are collectively expected to spend approximately $700 billion in Capital Expenditure in 2026, up from roughly $450 billion in 2025. Across the big five including Oracle (NYSE:ORCL), total hyperscaler capex in 2026 is forecast to exceed $600 billion, with approximately 75% directly tied to AI infrastructure.  This is the Demand foundation that makes Broadcom's $100 billion 2027 chip revenue target a trajectory, not a projection.

The decision by these organisations to direct a meaningful portion of that capital toward custom silicon rather than standardised GPUs reflects a structural judgment: that at their scale of compute consumption, purpose-built chips deliver superior unit Economics. Each of these companies has made multi-generational commitments to custom accelerator programmes. OpenAI and Anthropic, though earlier in their infrastructure build-out, are moving in the same direction.

When Broadcom reports $10.7 billion in AI semiconductor revenue for a single quarter, that number is the financial expression of those commitments. A beat tonight would imply the ramp is running ahead of schedule. An inline print confirms trajectory. A miss raises questions about whether hyperscaler capex plans are beginning to moderate. Any upgrade to the $100 billion 2027 target, or new customer-level disclosure, would carry significant weight with institutional investors.

Nvidia Remains Formidable, But the Market Is Watching Both

Any assessment of Broadcom's position requires an honest account of Nvidia Corporation (NASDAQ:NVDA). Nvidia commands approximately 80% of the AI accelerator market. Its Blackwell architecture is in heavy production deployment, and its Vera Rubin platform has entered full production. Its most recent data centre revenue of $75.2 billion for fiscal Q1 2027 grew 92% year over year. These are not the numbers of a company losing ground.

The distinction is architectural and strategic rather than competitive in the traditional sense. Nvidia serves the broadest possible market with the most flexible possible product. Broadcom serves a narrow set of the largest buyers with the most customised possible product. Both models are growing. The question is which generates more durable returns at the margin.

One data point is instructive. Nvidia trades at approximately 33 times earnings. Broadcom trades at 92.31 times earnings. The market is ascribing a significant premium to Broadcom's stickier revenue model. Whether that premium is justified depends substantially on what management says tonight.

The Valuation Question Cannot Be Deferred

Broadcom's stock has returned approximately 85% over the past year, reaching a 52-week high near $495 before today's modest pullback. At a P/E ratio of 92.31, the stock reflects expectations that are not modest. Forty-four of forty-seven analyst ratings are Buy, with price targets as high as $600.

That consensus is fragile in one specific way. A stock priced at 92 times earnings has no tolerance for ambiguity. A soft Q3 guidance number, a delay in any named customer ramp, or any signal that the AI capex cycle is entering a consolidation phase would test that conviction quickly. The broader semiconductor peer group is mixed today. Qualcomm Incorporated (NASDAQ:QCOM) and Texas Instruments Incorporated (NASDAQ:TXN) are modestly lower while Intel is outperforming. Broadcom's move is earnings-specific, not sector-driven.

What to Watch on the Call Tonight

Q3 AI semiconductor revenue guidance is the primary variable. Reaching $100 billion for fiscal 2027 requires a steep ramp in the second half, and any quantitative commentary on that trajectory will be parsed carefully. The sixth customer programme, which management flagged as deploying more than one gigawatt of capacity in 2027, has not been publicly named. Any disclosure there would be significant.

The Infrastructure Software segment, anchored by the $69 billion VMware Acquisition, generates Recurring Revenue and meaningful margin contribution. Renewal rates and enterprise migration commentary will provide a fuller picture beyond AI semiconductors.

Supply chain visibility on leading-edge wafer capacity, high-bandwidth memory, and substrate availability will matter for any investor modelling 2027 volumes with precision.

Conclusion

Broadcom arrives at tonight's earnings report having executed consistently across multiple quarters of accelerating AI demand. The custom silicon model is working. The hyperscaler relationships are deepening. And with approximately $700 billion in aggregate hyperscaler capex committed for 2026 alone, the demand runway is not in question.

The test tonight is not whether Broadcom is a strong business. It demonstrably is. The test is whether a business this strong can continue to grow into a valuation that has run well ahead of its current earnings base. At 92 times earnings near all-time highs, the margin for anything less than confirmation of the bull case is narrow.

Custom silicon is the direction the AI infrastructure market is moving. Broadcom is the dominant player in that transition. What remains open, and what tonight will begin to answer, is whether that dominance is already fully reflected in the price.