Shares of the iShares Semiconductor ETF (SOXX +0.34%) plunged 21.2% in July, according to data from S&P Global Market Intelligence.
The semiconductor sector exchange-traded fund holds most of the major chip stocks that benefit from the artificial intelligence build-out. Amid the agentic AI boom in the first half of 2026, the sector ETF rallied an astounding 112.8%.
Therefore, it's perhaps no surprise that the index experienced a pullback in July, despite the extremely strong results disclosed by top semiconductor companies throughout the month.
There were also some non-fundamental reasons for the ETF's decline in July, including bearish speculation on the AI build-out from Michael Burry, who is famous for calling the housing crash of 2008 years before it happened. In addition, we now know that a large and prominent AI-focused hedge fund, Situational Awareness, "blew up" in July after using too much leverage to bet on AI stocks. That likely caused "forced selling" in many AI stocks, exacerbating the decline.
NASDAQ: SOXX
Key Data Points
Burry, Kimi3, and Situational Awareness sink chips stocks
It should be noted that the semiconductor sector was absolutely ripe for a big sell-off heading into July. It's very rare for any individual stock to more than double in a year, let alone six months, and even rarer for a diversified stock index to do so.
This was perhaps not lost on Michael Burry, of The Big Short fame. Burry continues to be a short-seller and writes publicly about his positions. On June 30, he wrote on his Substack that he had shorted the SOXX ETF and several individual semiconductor stocks, including Nvidia (NVDA -0.10%) and memory maker Micron (MU -1.31%). The basic thesis is that AI stocks, after their huge run, were extremely overvalued. Burry cast doubt on the durability of the AI boom, comparing it to both the late-1990s dot-com boom and bust and the housing boom and bust a decade later. Even as the index sank throughout the month, Burry wrote updates saying that he had increased his short bets on these names, attributing a late-month rally to a "dead cat bounce."
Sentiment was further damaged following the July 16 release of the Kimi K3 open-source model from the Chinese AI lab Moonshot. The 2.7 trillion-parameter "open weights" model was shown to perform on par with or better than the current top models from U.S. labs, OpenAI and Anthropic, for certain applications.
One could argue that a massively capable, low-cost open-weight model would be a boon for AI infrastructure if it spurs increased demand. However, given the already negative sentiment on the AI trade, investors appeared to worry that OpenAI and Anthropic, along with other higher-cost U.S. models, would be forced to compete more on price. If that were the case, questions arose as to whether these labs would be able to honor their massive cloud computing commitments to the large U.S. hyperscalers, which are the largest buyers of chips.
Finally, the selling pressure was exacerbated by the forced sale and eventual liquidation of the public stock positions of the hedge fund Situational Awareness. Run by ex-OpenAI researcher Leopold Aschenbrenner. Situational Awareness had enjoyed a fantastic 400%-plus return in the first half of the year, with assets reportedly ballooning to $45 billion by the end of June. However, that outperformance reportedly came along with up to 400% leverage on SA's public market positions. Thus, when those positions reversed very quickly, the fund's margin loans were called, forcing the sale of positions and ultimately leading SA to sell its public stock portfolio to the hedge fund giant Citadel at a significant discount.
Image source: Getty Images.
Boom or doom?
Investors are obviously asking themselves whether July's sell-off is a mere pullback amid a continuing AI bull market or the start of something worse. It should be noted that there have been various AI "panics" over the past couple of years, including the introduction of DeepSeek in early 2025, the late-2025 panic over hyperscaler debt issuances, and the early 2026 panic over TurboQuant technology, which made memory use more efficient in AI inference. The AI build-out continued past all of these panics, and the stocks have marched higher.
That's not to say that things can't unravel as they did in past booms. However, it's also possible we are earlier in the AI boom than Burry and other skeptics let on. Strong recent semiconductor earnings and outlooks this month showed no signs of weakness at all in AI-related chip demand -- quite the opposite, actually, as demand only seems to be accelerating.
