AI chip expected gap sparks a semiconductor pullback: supply-demand mismatch and structural revaluation amid disruptions to Broadcom guidance

The semiconductor sector experienced a sharp shakeout in early June.
On June 5th, during pre-market trading, AI concept stocks such as Broadcom, Micron, AMD, and Nvidia faced collective sell-offs, further expanding losses in the semiconductor industry.
In the first week of June, the SOXX ETF saw a single-day decline of about 6%, and by June 9th, it dropped another roughly 10% to around $540, marking its worst single-day performance in recent years.
The trigger for this correction was very clear—Broadcom (AVGO)’s FY2026 AI chip revenue guidance failed to meet market expectations.

However, a deeper look at the data reveals that the market sell-off was driven more by a correction in expectations rather than a structural shift in AI chip demand.

Fundamentals of AI Chip Demand: Broadcom Data and the True Underlying Picture

Broadcom’s FY2026 Q2 earnings report released on June 3rd provides the most direct window into the market.
Total revenue for the quarter reached $22.2 billion, up 48% year-over-year, with an operating profit margin of a record 67%.
AI semiconductor revenue climbed to $10.8 billion, up 143% YoY, with AI semiconductor orders exceeding $30 billion in the quarter—significantly higher than the actual shipments of $10.8 billion.

Broadcom CEO Hock Tan’s comments during the earnings call are noteworthy: AI semiconductor demand is “simply insatiable.”
Looking ahead to Q3, the company expects AI semiconductor revenue to reach $16 billion, up over 200% YoY, and for FY2026, AI semiconductor revenue is projected at $56 billion, roughly 180% higher than the $20 billion in 2025.

Why is the market still selling off?
The core reason is that Broadcom’s full-year AI chip revenue guidance of $56 billion was below the market’s average expectation of $57.6 billion, a gap of about 2.8%, which triggered a sentiment correction.
Goldman Sachs later published a report offering a more comprehensive perspective: the firm revised its estimates for AI semiconductor revenue from FY2026 to FY2028 to $57 billion, $133 billion, and $193 billion respectively, noting that the slightly lower Q3 AI revenue guidance mainly reflects delays in ramp-up for new customers’ capacity, rather than demand deterioration.

Goldman emphasizes that Broadcom maintains its core view of AI semiconductor revenue “significantly exceeding $100 billion” in FY2027, supported by orders from six major custom chip customers—including Google, Meta, Anthropic, OpenAI, and two unnamed clients—with locked-in procurement orders totaling $6 billion.

Wall Street’s consensus indicates that the current market pricing discrepancy is not due to a disconnect between supply and demand fundamentals, but rather a misalignment between short-term expectations and long-term realities.
Goldman also highlights that Broadcom has secured all critical component supplies—covering storage, lasers, and packaging—necessary to support its revenue forecast for FY2027, while the overall supply chain tightness is increasing across the board.
In other words, the real constraint in AI chips is not on the demand side but on the supply side.

Semiconductor ETF and Stock Performance: Data from the Correction Phase

As of early June, the SOXX ETF (iShares Semiconductor ETF) has gained about 90% year-to-date, significantly outperforming the S&P 500’s roughly 22% return in the same period.
This means that even after the correction in the first half of June, the semiconductor sector still nearly doubled in the first half of 2026, with volatility characteristics similar to past growth bubble episodes—though with fundamental differences.

The ETF’s holdings have shifted noticeably.
By early June, the top three holdings were Micron, AMD, and Marvell, with Nvidia no longer the largest due to its 8% cap limit.
This structural change reflects a narrative shift from GPU compute power to storage and CPU inference bottlenecks, which is an intrinsic part of AI industry evolution.

BTIG strategist Krinsky’s assessment on June 8th indicated that SOXX still has about 14% to 17% downside potential to its 50-day moving average, which can be understood as a normal correction as the market digests excess gains.

Switching focus from ETFs to individual stocks, AMD closed at $511.57 on June 12th, up 4.73% for the day; Intel was at $124.57, up 6.51%; and Micron declined 1.43% to $981.61.
The divergence in gains and losses reflects sector rotation and the market’s differentiated pricing of various sub-sectors.

Revaluation of Storage Chips: Micron and the Strategic Upgrade of HBM

Amid the sector shakeout, the structural value of storage chips, especially high-bandwidth memory (HBM), is being redefined.
Traditional DRAM and NAND markets have long faced cyclical price volatility, but HBM—an essential component for AI accelerators—has fundamentally changed the pricing logic of this category.
Every Nvidia GPU requires HBM, and only three manufacturers globally can produce HBM at scale: SK Hynix, Samsung, and Micron.

Micron’s FY2026 Q2 results confirm this view.
Revenue for the quarter was $23.86 billion, up about 196% YoY, with non-GAAP EPS of $12.20, up 682%.
DRAM revenue reached $18.8 billion, NAND revenue $5 billion, reflecting strong demand from AI infrastructure build-out.
Non-GAAP gross margin expanded to 75%, and free cash flow reached $6.9 billion—exceeding Micron’s entire FY2024 revenue.

In terms of stock price, Micron hit a record high of $1,089.29 on June 3rd, with nearly 250% increase since the start of the year and a total return of 760% over the past 12 months, ranking among the top performers on Nasdaq.

Mainstream Wall Street institutions have set even higher targets.
Goldman Sachs raised Micron’s 12-month target from $400 to $900, and Wolfe Research further increased it to $1,250.
The core logic is that the supply-demand balance for DRAM and NAND is expected to remain tight through 2027, supporting sustained price and profit growth.

A noteworthy background is that Micron’s HBM capacity for 2026 is already sold out, providing a stable and predictable revenue base.
For its upcoming Q3 report on June 24th, the market expects about $34.4 billion in revenue, while Goldman’s forecast is $37.6 billion—about 9% above consensus, which, if realized, would further confirm the strategic upgrade of storage chips from “optional components” to “strategic assets” in the AI supply chain.

From Supply-Demand Mismatch to Trading Instruments: The Value of Gate Stock Contracts

In this environment, the key question for investors is how to effectively participate in AI chip trading.
Traditional brokerage account opening processes are cumbersome, with efficiency bottlenecks due to multiple account systems, and managing combined exposure to crypto assets and stocks is complex.
Gate’s stock trading feature, launched on June 12, 2026, offers a new solution.
The platform allows direct trading of major stocks and ETFs, including Micron and Samsung Electronics, using USDT, with perpetual contracts settled in USDT.

For traders focused on AI chip supply-demand logic, this tool offers three advantages:
First, using USDT as settlement means crypto funds can participate directly in stock markets without converting to fiat, reducing friction.
Second, perpetual contracts’ two-way mechanism enables investors to establish long positions during sector rallies and hedge via short positions during volatility.
Third, the platform has introduced leveraged ETF tokens, providing additional exposure options for different risk preferences.

Gate’s integration of crypto and stock assets into a single account system allows investors to manage cross-asset allocation and risk on one dashboard.

The Other Side of the Data: The True Mismatch in the AI Chip Market

Synthesizing the above data reveals several key structural features of the AI chip market.
Goldman Sachs’ on-the-ground survey of 24 core Asian supply chain companies in December 2025 indicates that demand for AI servers is expected to remain strong in 2026, with full-rack shipments potentially doubling, and ASIC chip shipments growing faster than GPUs.
Google’s TPU chips are among the highest-demand ASICs, and optical communication demand is booming due to upgrades from 800Gb to 1.6T speeds.

In contrast, the traditional terminal markets show a very different picture.
Goldman predicts that the PC market in 2026 will see minimal growth or slight decline, and smartphones face similar challenges—high-end models remain stable, but the low-end market is under pressure due to rising costs.
Automotive RF and analog chip demand remains weak.

This “AI boom, traditional weakness” duality means that the degree of re-pricing across the supply chain segments is uneven.
Companies in core AI sectors like HBM, ASIC custom chips, and optical communication are increasingly decoupling valuation logic from traditional semiconductor cycles;
Meanwhile, semiconductor companies relying on consumer electronics and automotive recovery are facing longer-term uncertainty.

Semiconductor wafer equipment spending is expected to grow steadily in 2026–2027, mainly driven by DRAM and advanced logic nodes, while NAND equipment spending remains subdued, with mature process logic under continued pressure.
This structural divergence in capital expenditure offers a reference framework for assessing the relative prospects of different sub-sectors over the next two to three years.

Conclusion

The semiconductor sector’s turbulence in the first half of June 2026 is essentially a short-term correction of overly high expectations.
The roughly 2.8% gap between Broadcom’s full-year guidance and market expectations, the nearly 90% gain in SOXX ETF in the first half, the technical reversion to the 50-day moving average, and BTIG strategist Krinsky’s estimate of 14%–17% additional downside—all form the current market pricing adjustment logic.
But all these adjustments are occurring within a macro backdrop of sustained growth in overall AI chip demand.
Micron’s full-year HBM capacity is sold out, Broadcom’s 2027 revenue target is backed by supply assurances, and Goldman’s tracking of strong growth in AI servers in 2026—all these fundamental signals have not reversed direction.

For investors, understanding the core of the AI chip market involves distinguishing short-term price signals from long-term structural trends.
The launch of Gate’s stock contract platform provides a real-time cross-asset trading tool, enabling investors to manage crypto and semiconductor stock exposures simultaneously on one platform—an integrated solution unavailable from traditional brokers.
Against the backdrop of ongoing battles between long-term AI infrastructure growth and short-term market corrections, every data point—from Broadcom’s guidance, SOXX’s trend, storage pricing, to ASIC capacity ramp-up—should be continuously monitored within a systemic supply-demand framework, rather than inferred linearly from single events.