The Philadelphia Semiconductor Index officially entered bear market territory on Friday, closing more than 20% below its late-June record high after falling as much as 5.7% during the session. The SOX had risen 105% between its March low and last month's peak. It has now given back more than a fifth of that gain in less than four weeks. The iShares Semiconductor ETF is down 18.6% in July alone, its worst monthly performance since 2008. Roughly $3.3 trillion in global semiconductor market value has been erased since June 22.
The catalyst Friday was Moonshot AI's unveiling of Kimi K3 at the World Artificial Intelligence Conference in Shanghai, a 2.8-trillion-parameter open-weight model that the company says rivals Anthropic's Fable and OpenAI's GPT-5.6. The launch was immediately dubbed "the Kimi moment" in trading desks across the US and Asia, drawing explicit comparisons to the DeepSeek shock in January 2026 that briefly erased nearly $600 billion from Nvidia's market cap. Compounding the pressure, Brent crude rose 3% to $86.73, raising fresh pass-through inflation risk at precisely the moment the Federal Reserve is already debating whether to hike in September.
Against that backdrop, the most practical question for individual investors is not whether the selloff was justified. It is what to actually do when a sector you own confirms a bear market while analysts and senior investors are publicly disagreeing about whether the fundamentals have changed.
Separate the Price From the Business
The single most important discipline in a momentum unwind of this scale is to distinguish between price movement and fundamental deterioration. They are not the same thing, and conflating them is the most common and most costly error individual investors make in periods like this one.
The SOX is in a bear market. That is a price fact. The underlying businesses are not in a bear market by any fundamental measure available today. Micron's contracted revenue base of $100 billion in take-or-pay agreements has not changed. BofA analyst Vivek Arya, writing Thursday, noted that spot prices for DRAM have climbed for eight consecutive weeks while NAND flash pricing has also strengthened. Memory spending as a share of hyperscaler capital expenditure has risen to 35 to 40% of total, two to three times the historical average, and Arya expects the wafer fab equipment market to reach $190 billion in 2027 and $250 billion in 2028. TSMC raised its full-year capital expenditure guidance from $52-56 billion to $60-64 billion this week, an $8 billion upward revision driven by customer demand, not speculation.
None of that changed on Friday when Kimi K3 was announced. What changed was sentiment about whether Chinese AI efficiency would eventually reduce the amount of compute hyperscalers need to buy. That is a legitimate question. It is a 2028 question, not a 2026 earnings question. The sell-side consensus EPS for Micron's fiscal 2027 of approximately $154 has not been cut. Forty-five analysts maintain a Strong Buy with an average target of $1,486.
The practical implication is that when you see a 20% sector decline and are trying to decide what to do, the first question is not "should I sell" or "should I buy more." It is "has anything changed about why I owned this in the first place." If the answer is no, the psychological weight of the price movement is doing your thinking for you rather than the analysis.
Understand What Is Actually Driving the Selloff
The current semiconductor bear market has at least four distinct drivers, and they have different implications for the durability of the decline.
The first is the momentum unwind. Goldman Sachs documented that the SOX momentum factor delivered its largest three-week drop relative to the S&P 500 since March 2001 before this week's additional declines. An 83% to 105% gain in three months creates a population of investors holding massive unrealized gains who are looking for permission to reduce risk. Every credible-sounding negative headline, from ASML's efficiency improvements to Kimi K3, provides that permission regardless of whether the fundamental thesis has changed.
The second is the Korean leveraged ETF dynamic. The assets under management of leveraged memory stock ETFs shrunk 34% from their June peak, and their AUM-to-underlying-market-cap ratio is three times higher than for equity ETFs broadly. That mechanical amplification of each down day created selling pressure disconnected from any fundamental view. South Korean regulators reversed their approval of those products this week, which removes the amplification mechanism going forward.
The third is the Kimi K3 narrative, which is the Friday-specific driver. The bear case is that cheap open-source Chinese AI compresses the revenue hyperscalers can earn on their AI capex, which in turn reduces the need for AI chips. The historical precedent from DeepSeek in January suggests this concern, while real, tends to be shorter-lived as an equity market driver than the initial reaction implies. DeepSeek triggered a $600 billion single-day wipeout in Nvidia; Nvidia subsequently recovered and reached new highs. The US maintains a six to nine month lead in frontier AI capability, according to most analysts, and Kimi K3 is being compared to models that Anthropic and OpenAI released several months ago.
The fourth is macro, specifically Brent at $86.73, the Fed's September hike probability remaining elevated, and the 10-year yield staying above 4.5%. Higher rates compress the multiples on high-PE growth stocks. This is the headwind that does not resolve with a single data point.
What BofA Is Actually Saying
Vivek Arya's characterization of this correction as "a summer reset, not a fundamental reversal" deserves to be unpacked rather than dismissed or accepted at face value. The seasonal data supports his framing: the SOX has underperformed the S&P 500 in the third quarter in 10 of the last 16 years. The current correction follows a quarter in which the index gained 80%. The historical pattern is that extreme first-half outperformance in semiconductor stocks is followed by summer underperformance that is painful to sit through but does not typically indicate the end of the structural cycle.
Arya's framework is not bullish because he is optimistic by disposition. It is bullish because the forward data he monitors, DRAM spot pricing up eight consecutive weeks, NAND flash strengthening, hyperscaler capex commitments unrevised, wafer fab equipment market on track for $190 billion in 2027, points in the same direction as the companies' own financial statements. A fundamental reversal requires those inputs to deteriorate. They have not.
That said, "not a fundamental reversal" is not the same as "the bottom is here." BofA's position is that the correction is seasonal and sentiment-driven, which implies it continues until the sentiment pressure exhausts itself, not until any specific price level is reached.
The Practical Framework for Right Now
Individual investors in the semiconductor sector are facing a situation where the price action is unambiguously bearish, the sentiment is near extreme fear readings, and the fundamental data has not confirmed the bear thesis. Historically, that combination resolves in one of two ways: the fundamental data eventually deteriorates to match the price, or the price eventually recovers to reflect the unchanged fundamentals.
The events that will begin to determine which scenario is unfolding are on a visible calendar. SK Hynix reports July 22. Micron, Samsung, and SK Hynix will all report before mid-August. The big tech earnings cycle provides hyperscaler capex commentary through early August. The July CPI prints in mid-August and will determine whether the Fed's inflation pressure is easing.
Gavin Baker, the technology-focused fund manager at Atreides Management, posted Wednesday that risk/reward has become attractive again in cheap stocks with durable competitive advantages. Bill Ackman agreed publicly. Neither is making a call on next week's price. Both are making a call on the 18 to 36 month earnings trajectory relative to current valuations. That is the relevant time frame for evaluating whether the current price level represents a dislocation or a correct re-rating.
The hardest thing to do in a bear market is to hold the analytical framework that drove a decision and evaluate it against new information rather than against the pain of the price decline. The easiest thing to do is to let the price movement do the thinking. The investors who consistently outperform over full cycles are the ones who know the difference.
Explore more exclusive insights at nextfin.ai.
