On June 3, 2026, Broadcom (NASDAQ: AVGO) released a Q2 earnings report that exceeded nearly every expectation, yet the market’s reaction was dramatic—shares plunged 12% after hours, dropped as much as 15% intraday the following day, and closed down 12.59%, wiping out over $32 billion in market value in a single day. This was not a flawed earnings report: total revenue for the quarter reached $22.19 billion, up 48% year-over-year and ahead of analysts’ forecast of $22.13 billion. Adjusted earnings per share came in at $2.44, beating the consensus estimate of $2.40. AI semiconductor revenue soared 143% year-over-year to a record $10.8 billion.
This was a textbook "expectation gap" trap. When a stock’s valuation already prices in every conceivable optimistic scenario, any information that fails to exceed expectations—even if it merely meets them—can trigger a valuation reset. Broadcom’s challenges didn’t end there. Almost simultaneously, Macquarie downgraded Broadcom from "Outperform" to "Neutral," slashing its target price from $513 to $437. The core rationale was Google’s accelerated push toward in-house chip design and diversified supplier strategies.
Behind the "Perfect Earnings": The Double Squeeze of Expectation Gap and Valuation Premium
Broadcom’s Q2 numbers were nearly flawless. Semiconductor solutions revenue hit $15.01 billion, up 79% year-over-year and beating the analyst consensus of $14.65 billion. Non-GAAP net income reached $12.07 billion, up 55%. Free cash flow totaled $10.26 billion, accounting for 46% of revenue. Infrastructure software business grew 9% year-over-year to $7.18 billion. CEO Hock Tan described AI demand as "simply insatiable" on the earnings call, confirming partnerships with six hyperscale clients including Anthropic, Google, Meta, and OpenAI, and announcing a collaboration with Apollo and Blackstone to build a 20-gigawatt AI compute platform.
Yet, the market’s focus landed on three metrics that failed to beat expectations.
Metric One: Q2 AI revenue fell short of buy-side expectations. Broadcom’s Q2 AI semiconductor revenue was $10.8 billion, topping the company’s own guidance of $10.7 billion, but missing buy-side forecasts of $11.3 billion.
Metric Two: Q3 AI guidance significantly below buy-side modeling. Broadcom projected Q3 AI chip revenue at $16 billion, up over 200% year-over-year, but this figure was about $1.2 billion shy of analysts’ models, which expected $17.2 billion. Broadcom’s stock had already rallied over 40% this quarter, and its valuation implied that "every guidance must set a new ceiling." The company’s guidance fell short of the market’s most aggressive expectations in quantitative terms.
Metric Three: Long-term AI targets unchanged. Hock Tan reiterated the long-term goal of AI semiconductor revenue "exceeding $100 billion" for fiscal year 2027, without raising it. This is above the consensus market forecast of $57.6 billion for 2026 AI revenue, but lags behind Goldman Sachs’ estimate of $125 billion. What worried the market even more was Broadcom’s conservative stance—unlike peers such as Marvell Technology, which issued optimistic long-term guidance, Broadcom’s approach heightened Wall Street’s concerns about long-term growth momentum.
For a stock that had already surged ahead of earnings and was trading at historic valuation highs, merely meeting expectations wasn’t enough—the market demanded "outperformance." Wedbush summed it up succinctly: "At a forward revenue multiple of 25-30x, the market needs not confirmation, but acceleration."
The macro environment is equally crucial. The day before the earnings release, the US Department of Labor reported April CPI data: headline CPI rose 3.8% year-over-year, above the 3.7% market expectation and up from 3.3% in March; core CPI rose 2.8%, also beating the 2.7% forecast. Sticky inflation, combined with energy costs driven by Israel-Iran geopolitical tensions, reinforced expectations that the Fed will maintain "higher for longer" interest rates. In a high-rate environment, high-valuation growth stocks are typically the first to come under pressure, and Broadcom’s rolling P/E ratio of about 87x before earnings made it one of the most sensitive targets.
Google’s In-House Chips: A Structural Turning Point for the ASIC Duopoly
If unchanged Q3 guidance and 2027 targets were disappointments in the short-term catalyst layer, Google’s in-house chip strategy poses a structural challenge to Broadcom’s ASIC business.
Google and Broadcom have collaborated on TPUs for over a decade. However, the previous single-sourcing model is being replaced by a systematic diversification strategy. Google’s custom chip supply chain now has three pillars: Broadcom handles high-performance TPU variants, MediaTek supplies cost-optimized "e" series TPUs (20%-30% lower cost), and TSMC manufactures them. MediaTek has aggressively entered the AI ASIC market in recent years; while official confirmation of Google orders is lacking, industry consensus holds that MediaTek has made substantial progress in securing TPU orders.
More impactful, market sources indicate Google has commissioned Marvell Technology to design custom network chips for its next-generation TPUs. These chips will use Intel’s advanced 18A or 18AP process, are slated for mass production by late 2027, and will be paired with MediaTek-designed Humufish TPUs. Network chips play a critical role in AI data center clusters, managing inter-chip data flow, congestion, synchronization, and latency—their strategic importance rivals that of the main compute chips. This isn’t Marvell’s first collaboration with Google in AI chips; since 2024, rumors have persisted that Google is actively pursuing Marvell for AI inference chip development, systematically diversifying its supply chain.
Macquarie analyst Arthur Lai quantified the impact in a recent report. The firm forecasts Broadcom’s share of Google TPU-related revenue will fall from about 95% in 2026 to 80% in 2027, and further to 65% by 2028. By this measure, Broadcom could lose tens of billions in ASIC market share from Google alone. Macquarie also cut its 2028 EPS forecast for Broadcom by 21%, warning that intensifying competition in the AI ASIC market could suppress Broadcom’s growth and profitability over the long term.
From Google’s perspective, the motivation for in-house and diversified chip development is clear. Broadcom charges royalties per TPU produced, and as TPU demand grows exponentially, Google’s payments to Broadcom have surged. Building its own chip capabilities and bringing in more external partners is Google’s inevitable path to controlling its AI compute supply chain.
Meanwhile, Marvell Technology is emerging as the other pole in the custom ASIC race. NVIDIA CEO Jensen Huang declared at Computex Taipei that Marvell is "the next trillion-dollar company," boosting market confidence in custom AI chips. Marvell’s data center custom chip business is now its fastest-growing segment, and its recent Q2 guidance centered on $2.7 billion in revenue, signaling a rapid expansion pace in the ASIC arena.
It’s important to note that Google’s "de-Broadcomization" isn’t a direct replacement, but a decentralization of the supply chain. Broadcom remains the primary design partner for high-performance TPUs, and its partnership with Google is renewed through 2031, keeping its revenue base relatively stable in the short to medium term. However, the structural trend is clear—hyperscale cloud providers are moving from relying on external ASIC suppliers for design to internalizing design and diversifying procurement across multiple vendors. This shift poses a significant long-term threat to Broadcom’s ASIC business moat.
AVGO Plunges 15%: A Data Panorama Amid Divergence
After the earnings report, Wall Street’s views on Broadcom diverged sharply, reflecting the market’s balancing act between short-term AI demand and long-term market share.
Sell-side divergence: Downgrades amid bullish camps.
Macquarie’s downgrade to "Neutral" and target price cut from $513 to $437 is the most prominent bearish signal. Morgan Stanley, Citi, and Wells Fargo maintained buy ratings, while Jefferies raised its target price to $550. The consensus analyst rating is "Moderate Buy," with average 12-month target prices ranging from $473 to $486—offering 11%-16% upside from the post-plunge closing price of $418.91.
Buy-side logic: Long-term bullishness isn’t based on short-term guidance.
Bullish institutions aren’t optimistic about Q3 guidance per se, but rely on two core premises. First, Goldman Sachs revised Broadcom’s AI revenue forecasts for 2026-2028 to $57 billion, $133 billion, and $193 billion, respectively, believing management could deliver over $40 billion in cumulative AI revenue by 2030. Second, as major clients’ XPUs ramp up in 2027-2028, Broadcom’s AI revenue growth inflection point will concentrate after 2027. Morningstar analysts, after the plunge, raised Broadcom’s fair value estimate from $550 to $650 per share, arguing the 18x consensus 2028 earnings multiple is attractive.
Insider trading signals.
Notably, Broadcom insiders sold about $356.4 million in stock over the past three months. While insider sales can have varied motivations and don’t necessarily reflect changes in company fundamentals, this data inevitably prompts extra scrutiny of management confidence during a sensitive earnings downturn.
Marvell’s dual-sided position.
Marvell Technology also experienced volatility in this wave—its shares dropped about 7% on earnings day, less than Broadcom, and partially rebounded the next day on news of Google’s network chip orders. Structurally, Google’s supply chain diversification is a net positive for Marvell: as the other half of the ASIC duopoly, Marvell’s share of Google chip design orders is rising. However, Marvell’s Q2 earnings showed quarterly revenue of only about $2 billion, while Broadcom’s AI business delivered $10.8 billion in a single quarter. Even with incremental Google orders, Marvell’s AI revenue scale remains an order of magnitude below Broadcom. Marvell’s potential gains are largely structural supplements, unlikely to become true competitive substitutes in the short term.
Conclusion
The ripple effects from Google’s in-house chip push aren’t Broadcom’s only short-term challenge. Looking at a longer time horizon, the core driver of the current plunge is the "expectation gap," not "deteriorating fundamentals." Broadcom’s AI business will only enter large-scale ramp-up after 2027, as XPU deployments from multiple hyperscale clients come online. The market is mainly digesting disappointment over the lack of an upward revision to the 2027 guidance ahead of time.
Macquarie’s $437 target price reflects pessimism about long-term market share loss, while the bullish camp’s $485-$550 targets signal sharply different long-term expectations. The fundamental debate centers on how quickly Google’s supply chain diversification and internalization will progress—is it gradual decentralization, or accelerated substitution?
For investors focused on the AI semiconductor theme, Broadcom’s plunge offers a window for observation. The key question is no longer whether Q3 guidance is strong enough, but how deep the custom ASIC business model’s moat remains amid the hyperscale clients’ internalization wave.
