Roughly $3.3 trillion in global semiconductor market value has been erased since June 22.
That is not a typo. That is one of the fastest sector drawdowns in modern market history, and most investors are still trying to figure out whether it is a correction or the beginning of something more structural.
Here is what we actually know right now. The Philadelphia Semiconductor Index fell more than 20% from its late-June record high last week, officially crossing into bear market territory. In a single week, the SOX dropped 11% — its worst weekly performance since March 2025. Before that, the same index had surged 105% between its March 2026 low and the June peak. The VanEck Semiconductor ETF (SMH) posted its third weekly decline in four weeks, falling nearly 9% in the most recent period.
And yet the sector is still up more than 60% in 2026. Intel shares are up over 160% year to date despite plunging more than 30% from the sector’s June peak. Marvell is up 121% on the year and still down nearly 40% from the SOX’s high. Both things are true simultaneously, and that tension is exactly what makes this the most interesting setup going into Tuesday’s earnings.
What Actually Broke the Trade
There were three triggers. Not one.
The first was valuation. After a 105% surge in three months, the sector had priced in a future that left zero margin for doubt. When Chinese AI startup Moonshot AI released a new model — Kimi K3 — that it claimed narrowed the gap with leading U.S. offerings, the market treated it as evidence that the compute-scarcity premium underpinning chip valuations had an expiration date. Investors had seen this movie before with DeepSeek. They did not wait around to find out if the sequel was different.
The second was Meta’s cloud announcement. After reports that Meta was exploring ways to monetize potential excess AI compute capacity, it challenged the most important premise behind the entire AI chip trade: that perpetual scarcity would keep prices and margins elevated. If the largest hyperscalers have excess capacity to sell, the supply-demand equation that justified extreme valuations begins to unwind.
The third was positioning. Funds tracking U.S. semiconductor stocks recorded outflows of roughly $11 billion in the week ended June 24 — the biggest weekly outflow this century. Short interest across the sector had reached a three-year high. When a crowded trade gets a narrative shift, the exits get narrow fast.
None of this means AI spending is slowing in any fundamental sense. The four major hyperscalers — Microsoft, Amazon, Google, and Meta — are still collectively guiding toward approximately $750 billion in combined AI capital spending for 2026, and estimates for 2027 approach $1 trillion. Memory now accounts for roughly 35% to 40% of AI cloud capital spending, more than double historical levels. The buildout is not reversing. What reversed was the multiple, not the demand.
Why Tuesday Is the Most Important Earnings Date in the Market Right Now
Alphabet reports July 22 alongside Tesla and GE Vernova. Of the three, Alphabet’s capex guidance is the single highest-stakes data point of the entire week — and possibly the season.
Here is why. If Alphabet signals any moderation in AI infrastructure spending, the ripple effect touches every chip company, every data center operator, every power infrastructure business, and every AI software company that depends on compute capacity. Conversely, maintained or expanded capex guidance would be the clearest possible signal that the semiconductor sell-off was a positioning reset rather than a demand story. Google Cloud has accelerated its growth rate across five consecutive quarters, reaching $20.0 billion in Q1. Cloud operating margin has expanded from 17.8% to 32.9% in a single year. The backlog in Q1 reached $462 billion in unrecognized revenue — nearly doubling in a quarter. Those numbers say the spending is working. The question Tuesday answers is whether it continues.
Intel reports on July 23. Texas Instruments reports the same evening as Alphabet and Tesla on July 22. Both matter for very different reasons.
Texas Instruments is the best analog-chip barometer in the market. TXN’s book-to-bill ratio has called every industrial cycle turn for a decade. With global manufacturing PMI at 52.5 — the strongest reading in years — TI either confirms the industrial recovery is real or starts to walk it back. Either answer has significant implications for the broader rotation trade into industrials and away from AI hardware.
Intel’s quarter is about something different entirely: foundry credibility. The company is up over 160% in 2026, making it one of the most dramatic AI-adjacent stories of the year. But the yield rates on its 18A process node and any commentary on customer wins will determine whether that run had legs underneath it or was purely a sentiment trade.
The Stocks Inside This Sell-off Worth Watching
Not all semiconductor names are in the same position. Nvidia trades near a forward price-to-earnings ratio that is at its lowest in more than ten years. Micron’s forward P/E touched a nine-year low earlier this year, and the company posted the largest quarterly revenue growth in semiconductor history. Bank of America has argued that memory is moving from a cyclical commodity to a strategic AI enabler, with memory now accounting for roughly 35% to 40% of AI cloud capital spending.
The stocks that fell hardest in the sell-off — Marvell, ARM, Intel — were also the ones that ran the most. ARM is still up significantly on the year despite a 30% drawdown from its peak. Marvell, at 121% year-to-date gains even after the correction, is one of the more striking examples of a sector that got dislocated from its own fundamentals on the way up and is now recalibrating on the way down.
Slight tangent, but worth noting: the retail sector quietly gained 2.5% in the same week the SOX dropped nearly 9%. Capital is rotating visibly. The S&P 500 financials sector logged back-to-back record closes during the chip sell-off. The Dow Transportation Average was up more than 30% on the year and near record territory. The rotation into real-economy sectors, financials, and energy is not a theory anymore. It is the price action of the past three weeks.
Technical Framework for the SOX
The SOX broke the neckline of a head-and-shoulders top formation in the most recent session, a technically significant event. The 12,000 level that had held twice as support gave way, and the index’s next downside technical target sits near 11,000 — roughly 7% below the Thursday close. The iShares Semiconductor ETF (SOXX) has already confirmed the bear market threshold with a close more than 20% below the June peak.
On the upside, a sustained move back above the 12,000 level in the SOX would require a catalyst — specifically, hyperscaler capex guidance that surprises to the upside on Tuesday. Without that, the technical picture suggests the path of least resistance is lower in the near term, even as the year-to-date gains remain substantial.
Volume patterns during the sell-off showed heavy institutional selling on down-days and limited buying on attempted rebounds. Goldman Sachs data showed hedge funds trimming technology-hardware exposure for a fourth consecutive week heading into the most recent sell-off. That is not the footprint of a contrarian buying opportunity yet — it is a trend in motion.
Three Scenarios for the Week Ahead
Bull Case
Alphabet reaffirms its 2026 AI capex commitment at or above $175 billion to $185 billion. Google Cloud growth accelerates beyond the 28% rate seen in recent quarters. Chip stocks stabilize, SOX finds support above 11,500, and the SOXX begins to recover toward the $560 level that marked pre-sell-off resistance. Marvell and AMD lead a relief rally, and the hyperscaler earnings season resets positive sentiment for the sector into August.
Base Case
Alphabet meets capex expectations without meaningfully raising them. Google Cloud grows in line with consensus. Semiconductor stocks stay rangebound near current levels as investors wait for Microsoft, Meta, and ultimately Nvidia (reporting August 26) before making larger moves. The sector remains in a consolidation pattern rather than entering a sustained recovery or a second leg lower.
Bear Case
Alphabet signals any pullback in AI infrastructure spending plans, or guides cloud revenue growth below the accelerating trajectory investors have come to expect. The SOX breaks decisively below 11,000. Marvell, Intel, and AMD each see renewed selling pressure. The narrative shifts from valuation reset to demand slowdown, and the sector gives back additional year-to-date gains through August.
Active Trader Framework
The most important discipline right now is separating the structural thesis from the positioning reality. The AI buildout is real. The demand for chips, power, cooling, and networking is compounding. But a sector that gained 105% in three months and then lost 20% in three weeks is telling you something specific: crowded trades require exceptional execution on every single catalyst. There is no cushion for disappointment at those entry prices.
Traders with long exposure to the semiconductor sector heading into Tuesday face a binary catalyst. The risk management question is not whether to own chips — it is whether the current position size is sized for the implied volatility of an event that could move the sector 5% to 10% in a single session in either direction.
For those looking at potential entry points, the selective approach makes more sense than broad sector exposure. Nvidia at a 10-plus year low on forward earnings is a different risk-reward than names still trading at elevated multiples relative to their own history. Micron, where Bank of America sees one of the biggest AI mispricings in the market, is a different conversation than ARM at 60x-plus forward earnings after a 30% correction.
The answer to whether this was a reset or a reversal lands Tuesday evening. And it will come from a company that does not make a single chip.
