Semiconductor selloff wipes out Samsung and SK Hynix as AI spending fears deepen

South Korea's Kospi index fell nearly 11% on Tuesday, triggering the country's eighth circuit breaker of 2026 and marking its weakest close since April. Samsung Electronics shed more than 13% and SK Hynix dropped over 14.7%, with SK Hynix's US-listed shares closing below their IPO price. The rout spread across the region and into US markets, where the Philadelphia Semiconductor Index fell as much as 6% across four consecutive losing sessions and the Nasdaq-100 declined 9.7% from its record high, putting the index within reach of correction territory.

The selloff didn't have a single trigger — it's the product of three separate concerns arriving at once, each reinforcing investor anxiety that had already been building over the scale of AI infrastructure spending.

China's Chipmaking Progress

The Information reported that a state-backed Chinese company has begun mass-producing the country's first homegrown immersion deep ultraviolet (DUV) lithography tools, with roughly five units expected to ship this year to chipmakers SMIC, Hua Hong, and CXMT, and about 20 more planned for 2027. Most components are domestic, though some critical parts still come from Japan, and supplier delays have held back output so far this year — meaning the tools are real but not yet at parity with leading-edge lithography equipment from ASML.

That report landed one day after Chinese memory chipmaker CXMT's debut on Shanghai's STAR Market, where shares surged 466% after the company raised $8.6 billion — a signal to investors that Chinese chip manufacturing capacity and capital access are both accelerating faster than expected, directly threatening the competitive position of Korean memory makers like Samsung and SK Hynix.

The AI Spending Sustainability Question

Layered on top of the China news is a growing unease about hyperscaler capital spending. Alphabet raised its 2026 capital expenditure guidance to $205 billion, up sharply from $91 billion in 2025, with 2027 spending expected to climb even higher — part of a projected $785 billion in combined 2026 spending across the six largest hyperscalers, according to Moody's, which noted that "the ultimate return on all of it is unclear." Alphabet posted its worst single trading day of the year following its earnings announcement, despite the capex increase reflecting continued confidence in AI demand rather than a pullback.

That uncertainty has been compounded by renewed scrutiny of circular financing arrangements in the AI supply chain. Nvidia is reportedly pursuing more than $750 billion in new deals, including a partnership with South Korea's SK Group worth over $500 billion in mutual business, and a proposed $250 billion financing guarantee to help OpenAI lease a planned Ohio data center campus. Critics argue that arrangements where a chipmaker finances its own customers' purchases of its chips can artificially inflate demand signals, magnifying losses across the industry if AI monetization doesn't keep pace with the spending.

Do the Fundamentals Support the Panic?

Several analysts have pushed back on reading this as a fundamentals-driven collapse. DRAM contract prices for the third quarter are reportedly settling 20% to 30% higher than the prior quarter, memory supply remains genuinely scarce as manufacturers redirect capacity toward AI-grade chips, and Google and Meta both have five-year supply agreements locked in with memory makers. Gil Luria, an analyst at DA Davidson, characterized the mood bluntly: "Right now there's a lot of panic around the AI investment, and the panic appears to be indiscriminate."

What This Means Going Forward

The gap between underlying demand fundamentals — which by most accounts remain strong — and the market's reaction suggests this selloff is being driven more by sentiment and circular-financing anxiety than by any actual deterioration in chip demand. That distinction matters for how long the rout lasts: sentiment-driven selloffs tend to reverse faster than ones rooted in genuine demand collapse, but they can also compound quickly if enough investors treat today's decline as confirmation of their existing AI-bubble concerns rather than a temporary overreaction, as detailed in Fortune's coverage of Tuesday's trading.