Chip stocks had their roughest stretch of 2026 in the first two weeks of July, and the sell-off had almost nothing to do with anyone actually wanting fewer AI chips. Micron led the drop with a single-session fall of roughly 13 percent that erased about $138 billion in market value in one day. Intel and AMD followed, down 9 and 7 percent in the same window, with Intel’s cumulative slide reaching more than 20 percent over the following days and AMD cratering a further 8 percent on a separate session tied to Samsung’s earnings release. Applied Materials, the equipment maker whose fortunes track the whole industry’s capital spending, fell 10 percent alongside them. Collectively, more than a trillion dollars in chip-sector market value evaporated in the space of roughly a week.
What makes the sell-off unusual is that it arrived during a period of genuinely strong underlying demand. Global semiconductor sales hit $120.6 billion in May 2026 alone, up 9.2 percent month over month and more than double the same month a year earlier. The catalysts behind the stock declines were narrower and more technical: reports that SK Hynix, fresh off pricing the largest foreign IPO in US history the same week, is quietly slowing the pace of its high-bandwidth memory production expansion; growing investor skepticism about whether the hundreds of billions of dollars hyperscalers are pouring into AI data centers will generate a return on anything close to the timeline currently priced into chip-stock valuations; and a Federal Reserve that has turned more hawkish than markets expected heading into the second half of the year.
The math behind the valuation anxiety is stark when laid out plainly. For Intel and AMD to genuinely grow into the stock prices they were trading at before the pullback, analysts estimate Intel would need to roughly quadruple its earnings beyond current 2026 projections, while AMD would need to nearly triple its own. Those aren’t modest, multi-year-plan numbers, they’re the kind of growth that requires everything in the AI infrastructure buildout to go right, on schedule, with no meaningful pricing pressure from competitors or customers along the way.
Some context matters here, because a trillion-dollar sell-off sounds more alarming in isolation than it is against the run-up that preceded it. Intel’s stock had roughly tripled since the start of 2026, and AMD was up around 140 percent over the same period, both driven by the same AI infrastructure narrative that’s now under scrutiny. A pullback of this size, sharp as it was, mostly pauses those gains rather than erasing them; investors who bought in January are still sitting on substantial profits even after the worst week of the year.
Samsung’s own earnings, released in the middle of the sell-off, illustrate how confusing the signal has become. Samsung reported a nineteen-fold jump in quarterly operating profit, driven by the same memory shortage benefiting SK Hynix, and rather than reassuring the market, that number appears to have spooked it further, read by some investors as evidence the memory pricing cycle might already be near its peak rather than in its early innings. A blowout quarter triggering a sector-wide sell-off is not the behavior of a market confident in the multi-year story it was pricing in just weeks earlier.
CNBC’s own framing of the moment captured the whiplash well: chip stocks that had just notched record rallies through the second quarter of 2026 opened the third quarter with what the network bluntly called a dud. That kind of immediate reversal, from record highs to a trillion-dollar rout inside the same earnings season, is itself a signal worth taking seriously regardless of which side of the AI-infrastructure-ROI debate eventually proves right: it means the market’s conviction in the chip trade’s story is currently thinner and more reactive than the size of the preceding rally suggested.
It’s also worth being precise about what this sell-off is not. Unlike the speculative traffic that inflated dot-com-era valuations with little revenue behind them, the products at the center of this correction, HBM, AI accelerators, advanced packaging, all carry real order backlogs and contracted revenue today. The debate roiling chip stocks isn’t whether AI infrastructure spending is real, it’s whether the pace and scale of that spending can keep generating returns on the current investment timeline, a much narrower and more legitimate question than the one dot-com investors were dodging in 2000.
For the Philippines, the relevant exposure runs through the export economy rather than the stock market. Semiconductors remain the country’s single largest export category, and the electronics assembly and testing plants clustered around Cavite, Laguna, and Clark sit directly downstream of the same capital spending cycle that’s now under scrutiny on Wall Street. A slowdown in hyperscaler AI infrastructure investment, should the valuation skeptics prove right, would eventually show up as softer order volumes for Philippine test-and-assembly operations well before it shows up in any local headline. Filipino founders building AI-adjacent products on the assumption that compute costs will simply keep falling on a predictable schedule should treat this volatility as a reminder that even the picks-and-shovels layer of the AI trade is not immune to sudden, sentiment-driven repricing.
Share this article