Broadcom's Record Quarter Masks a Market Hungry for More

Broadcom delivered record Q2 revenue of $22.2 billion, yet a brutal after-hours sell-off exposed just how unforgiving expectations have become for the sector’s defining infrastructure play.

Key Takeaways

• Broadcom reported record Q2 2026 revenue of $22.2 billion, up 48% year-over-year, with AI semiconductor sales hitting $10.8 billion, a 143% surge that now represents nearly half of total company revenue.
• Despite beating consensus on earnings per share and free cash flow of $10.3 billion, shares fell more than 13% after hours, erasing approximately $250 billion in market capitalization in a matter of minutes.
• Q3 guidance of $29.4 billion in revenue, with AI chip sales projected above $16 billion, reflects contracted multi-year design wins that give Broadcom a visibility profile virtually unmatched anywhere in the semiconductor industry.

A Record Quarter Meets an Unmerciful Bar

Numbers of this magnitude rarely arrive quietly. When Broadcom reported its fiscal second-quarter 2026 results after the close on June 3, the financial evidence was categorical: consolidated revenue of $22.187 billion, up 48 percent from $15.004 billion in the year-prior period; AI semiconductor sales of $10.8 billion, up 143 percent; adjusted EBITDA of $15.244 billion, representing a 69 percent margin; free cash flow of $10.262 billion, up 60 percent; and non-GAAP diluted earnings per share of $2.44, comfortably clearing consensus. GAAP net income more than doubled to $9.310 billion. By any conventional measure, the quarter was a statement of intent.

The market’s reply was swift and severe. Within minutes of the release, Broadcom Inc. (NASDAQ: AVGO) shares fell more than 13 percent in after-hours trading, erasing roughly $250 billion in market capitalization from a pre-announcement close near $479. The scale of that reaction, set against results of this quality, defines the central tension now embedded in Broadcom’s valuation. This was not a company being punished for failure. It was a company experiencing what happens when sustained outperformance becomes the baseline expectation rather than the exception, and when even a record quarter arrives without the upward revision that a premium multiple demands as its price of admission.

The Financial Architecture of an AI Infrastructure Leader

The composition of Broadcom’s revenue tells a more precise story than the headline figure alone. Semiconductor Solutions, the division encompassing both custom AI accelerators and AI networking silicon, generated $15.009 billion in the quarter, up 79 percent year-over-year. Its share of total company revenue has expanded from 56 percent to 68 percent in twelve months, a structural shift that reflects the deliberate repositioning of the business around the most capital-intensive phase of the AI buildout. Within that segment, the $10.8 billion in AI-related sales now represents nearly half of Broadcom’s total revenue, a concentration that would have seemed improbable as recently as two years ago.

The profitability profile reinforces the quality of that growth. Adjusted EBITDA margins of 69 percent on a revenue base approaching $22.2 billion are a function of both operating leverage and the pricing authority that comes with deep, multi-year customer integration. Free cash flow conversion of $10.262 billion, up 60 percent, means the company is generating capital at a pace that supports both sustained investment and consistent shareholder returns simultaneously. Non-GAAP diluted EPS of $2.44 represents 54 percent growth, and GAAP net income of $9.310 billion more than doubled, underscoring that the profitability gains are not a product of accounting adjustments but of genuine operating expansion.

The infrastructure software segment, anchored by VMware following its acquisition, contributed $7.178 billion, up 9 percent. While that growth rate trails the semiconductor side and disappointed investors hoping for a more visible acceleration in the post-integration period, the segment’s financial role is structural rather than promotional. It delivers high-margin, recurring revenue with limited capital intensity and minimal cyclical exposure, providing the cash-flow stability that funds dividends and buybacks without competing for the reinvestment demands of the hardware business. The two segments are not in tension; they are complementary in a way that gives Broadcom a financial resilience that pure-play AI hardware companies cannot replicate.

Custom Silicon and the Networking Layer

The source of Broadcom’s competitive position in AI infrastructure is neither accidental nor easily replicated. The company has built, over several years, a capability set specifically suited to the moment hyperscalers are now navigating: the transition from general-purpose GPU clusters to custom, workload-optimized silicon designed around each customer’s software stack, power envelope, and cluster architecture. Custom application-specific integrated circuits for Google’s Tensor Processing Units, Meta’s MTIA, Microsoft’s Maia, and programs associated with OpenAI and Anthropic have moved from early design collaboration to volume production, generating the revenue step-change visible in the Q2 figures.

What enables that transition is Broadcom’s advanced packaging capability. The company’s February 2026 introduction of a 2-nanometer custom compute system-on-chip built on a 3.5D XDSiP platform, integrating large silicon areas with multiple high-bandwidth memory stacks, delivers the performance-per-watt profile that makes custom silicon economically compelling at the scale hyperscalers now require. Alongside the silicon itself, AI networking, which contributed materially to the $10.8 billion AI revenue figure, supplies the Ethernet-based interconnect infrastructure that allows these custom clusters to function at tens or hundreds of thousands of accelerators without becoming constrained by inter-chip communication. At that scale, networking is not infrastructure overhead; it is the architecture that makes the entire system viable.

Guidance That Reframes the Trajectory

If the quarterly results confirmed Broadcom’s current position, the forward guidance reframed the scale of what is coming. For the fiscal third quarter ending August 2, 2026, management guided consolidated revenue to approximately $29.4 billion, representing 84 percent year-over-year growth, with non-GAAP operating margin holding near 67 percent. AI semiconductor revenue alone is expected to exceed $16 billion in the quarter, implying growth of more than 200 percent on the comparable prior-year period.

The credibility of that guidance rests on a structural feature of the custom ASIC business that distinguishes it from most of the semiconductor sector. Design wins are not spot-market transactions; they are multi-year commitments, locked in at the architecture stage, that typically extend visibility through 2028 and beyond. Broadcom’s management has referenced a substantial contracted AI backlog underpinning the ramp, and the Q3 figure is consistent with a second half of fiscal 2026 in which additional custom programs and next-generation networking silicon enter volume production simultaneously. That combination of contracted backlog and converging product cycles gives the guidance a durability that short-cycle semiconductor forecasts rarely possess.

What the Sell-Off Reveals, and What It Does Not

The 13 percent after-hours decline demands an honest reading, separate from the instinct to dismiss it as irrational. Several genuine concerns converged on June 3. Software growth, while financially sound, did not accelerate visibly enough to indicate the VMware integration entering a new, higher-velocity phase. The increasing concentration of revenue in AI semiconductors, a sign of strategic success, also introduces margin variability that investors in a premium-multiple stock will price with discipline. And management did not expand the longer-term AI revenue trajectory beyond the path already communicated, a decision that in a market scrutinizing every signal of AI capital expenditure durability can be interpreted as caution even when it is simply discipline.

What the sell-off does not reveal is any credible challenge to the fundamental investment case. Broadcom closed the quarter with $19.6 billion in cash and equivalents, returned capital through a $0.65 per-share quarterly dividend and continued share repurchases, and maintained a customer base spanning the major cloud builders and the leading frontier AI laboratories, a breadth that materially reduces single-customer concentration risk. The financial engine is operating at a pace, and with a margin profile, that few technology companies at this revenue scale have sustained.

The June 3 results captured a company whose AI strategy is scaling faster than most observers anticipated even twelve months ago. The market’s reaction captured something equally important: the standard applied to that company has scaled alongside it. For investors with the orientation and patience to look past a single session’s price action, the underlying evidence points in one direction.