Chip stocks tumble into bear market as AI rally unravels

The AI trade had a good run. A really good run. From March to late June 2026, the Philadelphia Semiconductor Index (SOX) surged 105%, powered by insatiable demand for memory chips and AI accelerators. Then came July, and the bill arrived.

The SOX has now fallen more than 20% from its late June high, the threshold that officially confirms a bear market. In a single week, the index dropped 11%, its worst weekly performance since March 2025. That is not a dip. That is a reckoning.

What went up, came down

To understand how we got here, you have to appreciate just how vertical the preceding rally was. A 105% gain in roughly three months is not normal price discovery. It is a market on a sugar rush, pricing in a future that may arrive much later, or differently, than investors assumed.

Memory chip stocks led the charge on the way up, and names like Kioxia and SanDisk were still sitting on gains of close to 600% year to date even after the recent pullbacks. In English: even after losing a fifth of their value, some of these stocks were still up six times from January.

The unwinding reflects a few overlapping anxieties. First, profit-taking was inevitable after a move that dramatic. Second, and more structurally, investors are beginning to question whether the AI capital expenditure supercycle can sustain the kind of chip demand that justified those prices. Hyperscalers have been spending aggressively, but the returns on that infrastructure spending remain, for now, more promise than proof.

The broader macro backdrop did not help. South Korea’s KOSPI fell nearly 20% from its own peak in just two weeks, entering bear market territory alongside the SOX. Japan’s Nikkei also joined the crowd. When multiple major indices break down simultaneously, it tends to accelerate selling pressure rather than contain it, as risk managers cut exposure across correlated assets.

Crypto felt the tremors too

Here is where it gets interesting for digital asset investors. Bitcoin fell to around $63,000 during the semiconductor sell-off, with analysts watching support in the $59,000 to $60,000 range below. That is not a coincidence.

The correlation between high-beta tech stocks and crypto is well-documented at this point. Both asset classes attract the same risk-on capital. When institutional investors and hedge funds need to reduce exposure quickly, they sell what they can, and liquid risk assets, both chips and coins, get hit in tandem.

The so-called AI token cohort felt it acutely. Tokens like FET, RENDER, TAO, and AGIX all showed significant price swings during the volatility. These assets are doubly exposed: they track crypto market sentiment broadly, and they carry a specific AI narrative premium that deflates when the underlying AI equity trade rolls over.

Think of it this way. If investors are suddenly skeptical that Nvidia’s next-generation chips will generate the returns Wall Street modeled, they are going to be even more skeptical that a crypto token representing distributed AI compute will generate those returns. The AI narrative is a single tent. When the poles wobble, everything underneath feels it.

The unprofitable tech index, a basket of loss-making growth companies, also slipped below its 50-day moving average for the first time in months, a technical signal that traders monitor closely as a leading indicator of broader risk appetite.

What this means for investors navigating both worlds

The semiconductor bear market is functioning as a stress test for the AI investment thesis writ large. For the past year-plus, the market narrative has been relatively simple: AI demand is insatiable, chip supply is constrained, buy anything in the stack. That trade worked spectacularly, right up until it did not.

The reassessment now underway is less about whether AI is real and more about timing and valuation. A technology can be genuinely transformative and still produce a bubble in the stocks tied to it. Those two things are not mutually exclusive. Investors who conflated the two may be discovering that distinction now.

For crypto-native investors, the read-through is specific. AI tokens were among the strongest performers during the semiconductor rally, attracting capital on the thesis that decentralized AI infrastructure would capture a share of the same demand wave driving Nvidia and its peers. That thesis has not been disproven, but the near-term price action reflects that it was priced generously.

Risk management is the operative phrase. The correlation between SOX and crypto markets means that a continued drawdown in semiconductors would likely sustain pressure on digital assets broadly, and on AI tokens specifically. Bitcoin’s proximity to the $59,000 to $60,000 support band is a level worth watching, not because a break there would be catastrophic, but because it would signal that the risk-off impulse has more room to run.

Investors with exposure across both traditional AI equities and crypto should be stress-testing their portfolios against a scenario where the AI capex narrative faces a more prolonged skeptical period, one measured in quarters rather than weeks. Kioxia and SanDisk being up 600% year to date is a reminder of how much froth accumulated. Even after a 20% correction in the index, the process of normalization may not be complete.

Disclosure: This article was edited by Estefano Gomez. For more information on how we create and review content, see our Editorial Policy.