On June 3, 2026, Broadcom Inc. (NASDAQ: AVGO) surged 5.2% in afternoon trading to a then-intraday record, closing at a new all-time high above $481 \u2014 its 52-week high having reached $488.82. That move pushed the company’s market capitalisation past $2.1 trillion, cementing its position as one of the most valuable semiconductor companies on earth. This was before a single number from its fiscal Q2 2026 earnings had been disclosed to the public.<\/p>\n\n\n\n
What drove the intraday surge was a convergence of five distinct catalysts, each independently bullish for Broadcom’s AI chip business. What happened after the earnings were released \u2014 a sharp 12% after-hours decline that erased more than $300 billion in market value \u2014 turned a straightforward triumph into a more nuanced story about the gap between a genuinely record-breaking quarter and a market that had priced in perfection.<\/p>\n\n\n\n
Understanding both moves is essential for investors trying to evaluate what Broadcom actually is in 2026: not a chip company, and not merely a software company, but the infrastructure backbone of the global AI buildout.<\/p>\n\n\n\n
Catalyst One: Alphabet’s $80 Billion AI Capital Deployment<\/strong><\/p>\n\n\n\n The most direct signal came from Alphabet, which announced plans to raise $80 billion \u2014 including a $10 billion investment from Berkshire Hathaway \u2014 earmarked for AI computing infrastructure. Broadcom’s relationship with Alphabet is structural: it designs Google’s Tensor Processing Units (TPUs) under a supply agreement that runs through 2031. When Google’s parent commits $80 billion to AI infrastructure, Broadcom is, by contract, a direct beneficiary. Markets made that connection within minutes of the Alphabet announcement.<\/p>\n\n\n\n Catalyst Two: Anthropic’s Confidential IPO Filing<\/strong><\/p>\n\n\n\n On the same day, Anthropic \u2014 one of Broadcom’s disclosed XPU customers \u2014 filed its draft registration statement with the SEC. Anthropic’s public listing at a near-trillion-dollar valuation raised the visibility of large-scale AI infrastructure commitments: the company has publicly committed to potentially 1 million TPUs from Google, the chips Broadcom co-designs. The IPO filing reinforced investor conviction that Broadcom’s hyperscaler customers have multi-year demand visibility, not just current quarter demand.<\/p>\n\n\n\n Catalyst Three: Jensen Huang’s Endorsement of the Custom ASIC Ecosystem<\/strong><\/p>\n\n\n\n At Computex, Nvidia CEO Jensen Huang endorsed Marvell \u2014 Broadcom’s closest rival in custom AI silicon \u2014 as the “next trillion-dollar company.” This validation of the custom ASIC ecosystem, coming from the dominant force in AI chips, implicitly affirmed the entire market in which Broadcom holds approximately 60\u201370% share. If the second-place player is worth a trillion dollars, the market leader is worth considerably more.<\/p>\n\n\n\n Catalyst Four: Rising Analyst Price Targets<\/strong><\/p>\n\n\n\n Into the earnings, Wolfe Research and Citi both raised their price targets on Broadcom to $500. Deutsche Bank analyst Ross Seymore had raised his target to $515. Susquehanna’s Christopher Rolland had earlier set a $450 target tied specifically to the Google TPUv7 chip cycle performing above expectations. The consensus was unusually aligned: strong buy, record targets, accelerating AI revenue visibility.<\/p>\n\n\n\n Catalyst Five: Pre-Earnings Positioning<\/strong><\/p>\n\n\n\n The combination of the above meant institutional investors were actively building positions ahead of a report they expected to be strong. The 5.2% afternoon surge was, in part, a function of that positioning \u2014 the market pricing in outcomes before they were confirmed.<\/p>\n\n\n\n Broadcom reported its fiscal Q2 2026 results after the close on June 3, 2026, for the quarter ended May 3, 2026.<\/p>\n\n\n\n The AI semiconductor numbers were genuinely exceptional. Revenue from AI chips reached $10.8 billion in the quarter \u2014 up 143% year-over-year and above management’s own guidance of $10.7 billion. That figure includes both custom AI accelerators (XPUs) designed for hyperscalers and AI networking revenue from Broadcom’s Ethernet switching and routing products. Together, AI now represents approximately half of Broadcom’s total semiconductor revenue.<\/p>\n\n\n\n Total consolidated revenue was $22.19 billion \u2014 up 48% year-over-year and a company record. It narrowly beat the Wall Street consensus of $22.04 billion. Adjusted EBITDA reached $15.2 billion, representing 69% of revenue \u2014 also a record \u2014 and up 52% year-on-year. Non-GAAP earnings per share of $2.44 beat the analyst consensus of $2.39 by $0.05. Free cash flow reached $10.26 billion, representing 46% of total revenue \u2014 a record. The company declared a quarterly dividend of $0.65 per share, payable June 30, 2026.<\/p>\n\n\n\n The Q3 FY2026 guidance was where the markets paused. Broadcom guided Q3 consolidated revenue of $29.4 billion \u2014 an 84% year-over-year increase \u2014 with non-GAAP operating margins stable at 67%. Within that, AI semiconductor revenue for Q3 was guided at $16 billion, implying over 200% year-on-year growth. Full-year FY2026 AI revenue guidance was reaffirmed at approximately $56 billion, representing approximately 180% growth for the full year. Management reiterated its FY2027 target of AI chip revenue exceeding $100 billion.<\/p>\n\n\n\n The after-hours reaction \u2014 a decline of approximately 12%, erasing over $300 billion in market value \u2014 is one of the more instructive examples of how markets can simultaneously process excellent absolute results and disappointing relative ones.<\/p>\n\n\n\n Two factors drove the post-earnings decline. First, Broadcom’s infrastructure software segment \u2014 which includes the acquired VMware business \u2014 generated $7.18 billion in revenue, missing the analyst consensus of $7.32 billion. VMware had been the steady, high-margin engine generating 78% margins and providing the cash flow that funds AI expansion. A $140 million miss against a $7.32 billion consensus is less than 2% off, but in a stock that had rallied to reflect perfection, any shortfall triggered profit-taking.<\/p>\n\n\n\n Second, and more significantly, Bernstein analyst Stacy Rasgon noted that the forward AI guidance, while directionally strong, fell short of what more aggressive buy-side models had pencilled in for Q3. Some institutional investors had modelled Q3 AI revenue above $16 billion; management’s $16 billion figure was exactly in line with the consensus but not above it. In a stock trading at a premium valuation that implies continuous upside surprises, meeting guidance precisely is not enough to sustain the pre-earnings premium.<\/p>\n\n\n\n The disconnect is a recurring feature of high-momentum growth stocks: the quality of the quarter was beyond dispute. AI revenue grew 143%. EPS beat. Free cash flow hit a record. The issue was that the stock had priced in more than a record quarter \u2014 it had priced in a quarter so strong that future guidance would also need to be raised above consensus.<\/p>\n\n\n\n Beneath the earnings volatility is a business model that is genuinely unlike anything else in the semiconductor industry, and understanding it is essential context for long-term investors.<\/p>\n\n\n\n Broadcom’s AI chip business is built on what the company calls XPUs \u2014 custom Application-Specific Integrated Circuits co-designed directly with each hyperscaler customer. Unlike Nvidia’s general-purpose H100 and B200 GPUs, which any buyer can purchase, Broadcom’s XPUs are purpose-built for one customer’s specific AI model architecture, fabricated at TSMC’s 3-nanometer node over an 18-to-24-month design cycle. The consequence is a switching cost that is close to absolute: Google cannot simply cancel its TPU programme and buy Nvidia GPUs the next quarter without losing years of architectural optimisation. The relationship is embedded.<\/p>\n\n\n\n Broadcom has confirmed six major XPU customers in volume production or committed ramps. Publicly named customers include Google, Meta, and Anthropic. In December 2025, CEO Hock Tan disclosed that a then-unidentified customer had placed a $10 billion chip order \u2014 widely reported to be OpenAI \u2014 one of the largest single-customer commitments in semiconductor history. The total AI backlog as of the Q2 report stood at approximately $73 billion. Hock Tan’s own compensation package, structured around 2030 targets, includes achieving AI revenue above $120 billion \u2014 a target that implies the company expects its AI business to grow roughly sixfold from the FY2025 level of approximately $20 billion over five years.<\/p>\n\n\n\n The networking component of AI revenue \u2014 Broadcom’s Ethernet switching, routing, and connectivity products \u2014 provides a second layer of AI exposure that does not require winning specific XPU contracts. Every AI data centre, regardless of whether it uses Broadcom or Nvidia accelerators for compute, requires networking. Broadcom’s Tomahawk switching chips and Jericho routing platforms are the dominant infrastructure layer for AI cluster interconnect.<\/p>\n\n\n\n The bear case is not difficult to articulate, and it is the same thesis that drove the post-earnings decline. AI chip revenue is now approximately half of Broadcom’s total revenue, and AI infrastructure is concentrated among a handful of hyperscaler customers. Any slowdown in that spending \u2014 whether from regulatory friction, economic softening, or a shift in AI architecture that reduces XPU relevance \u2014 would hit Broadcom’s revenue and stock simultaneously. The software miss in Q2 is a reminder that the VMware engine, which provides the margin floor, is also not immune to competitive pressure and integration risk.<\/p>\n\n\n\n The valuation question is real. Even after the post-earnings decline, AVGO trades at a premium multiple \u2014 PEG ratio of 0.63 according to Investing.com \u2014 which is actually modest relative to the growth trajectory, but represents a high absolute earnings multiple that leaves limited room for negative surprises.<\/p>\n\n\n\n Against that, $10.26 billion in free cash flow in a single quarter, a $73 billion AI backlog, multi-year supply agreements with the world’s largest technology companies, and a management team that has met or beaten AI guidance in every quarter since the programme began are not trivial offsets.<\/p>\n\n\n\n What June 3, 2026 illustrated was the distance between Broadcom’s operational performance \u2014 which was genuinely extraordinary \u2014 and the increasingly elevated expectations that its own stock price creates. The former continues to compound. The latter requires constant outperformance to sustain.<\/p>\n\n\n\nWhat the Actual Q2 FY2026 Numbers Showed<\/h2>\n\n\n\n
Why the Stock Fell After a Record Quarter<\/h2>\n\n\n\n
The XPU Moat: Why Broadcom’s Competitive Position Is Structurally Different<\/h2>\n\n\n\n
The Bear Case and the Balance<\/h2>\n\n\n\n