More than $1.3 trillion has been stripped from the world's largest semiconductor stocks in a matter of days, as two forces collided: a reported Chinese breakthrough in chip manufacturing equipment and deepening unease about whether the colossal sums being poured into artificial intelligence will ever generate a sufficient return. The sell-off has been global in scope, sweeping from Seoul to Tokyo, Amsterdam, and New York.

Seoul at the epicentre

South Korea's benchmark KOSPI index has fallen roughly 29% over the past month, entering bear market territory, with the country's two dominant chipmakers bearing the brunt. SK Hynix and Samsung Electronics — among the world's largest suppliers of the high-bandwidth memory chips that power AI servers — have seen their shares fall sharply. On one particularly bruising session this week, SK Hynix closed down nearly 15% and Samsung lost more than 13%, wiping billions from their market values in a single day.

The contagion spread quickly. Japan's Kioxia plunged more than 18%, Tokyo Electron dropped nearly 11%, and Taiwan's TSMC closed almost 3% lower. In Europe, shares in ASML, the Dutch company that makes the machines used to manufacture advanced chips, fell sharply. In the US, Micron, Intel and AMD extended losses across multiple sessions. CNBC reported that 20 of the world's most valuable chip companies lost $1.3 trillion in market value since last Friday's close, according to FactSet data.

"This decline appears to be driven largely by sentiment rather than fundamentals" — Michael Field, chief equity strategist at Morningstar

China's lithography advance rattles the industry

One immediate trigger for the sell-off was a report in The Information that a state-backed company in Shanghai has begun producing domestic immersion deep ultraviolet lithography machines — the tightly controlled equipment used to stamp circuit patterns onto silicon wafers. Reuters confirmed the development. The machines are expected to be delivered this year to major Chinese chipmakers SMIC, Hua Hong Semiconductor, and memory producer ChangXin Memory Technologies. Initial volumes are modest: five units planned for 2026, rising to around 20 in 2027. By comparison, ASML shipped approximately 131 immersion systems in 2025 alone, commanding a 98.7% share of the global market for such tools.

Analysts cautioned against overstating the immediate competitive threat. The Chinese machine matches roughly the specifications of ASML equipment first shipped around 2008, and some critical components still rely on Japanese suppliers. CNBC quoted a team of analysts at SemiAnalysis saying that "tool performance, scaling production of the machine itself, fleet performance, surrounding ecosystem" all work against the Chinese DUV as a near-term substitute. What the development does provide, however, is a domestic supply path that cannot be severed by future Western export controls — a strategically significant shift even if it is not a capability leap.

The bigger question: who pays for the AI buildout?

Beneath the China story lies a more fundamental anxiety. The four largest US tech companies — Amazon, Alphabet, Meta, and Microsoft — plan to spend a combined $725 billion on AI capital expenditure in 2026, up 77% from the previous year's already record $410 billion, according to multiple analyst estimates. Aggregate capital spending by the world's largest hyperscalers is on track to overtake their operating cash flow around the third quarter of this year, according to research firm Epoch AI, meaning most have already turned to external financing to fund their buildouts.

Meta has raised its full-year capital expenditure guidance to between $125 billion and $145 billion, the largest single-year increase among its peers. Alphabet, which has doubled its long-term debt to $98.2 billion to fund its infrastructure push, saw its chief financial officer warn that free cash flow will remain under pressure. Investor reaction has grown increasingly divided: when Meta and Microsoft announced their aggressive spending plans, their shares fell, whereas Alphabet and Amazon rose after reporting cloud revenue growth that suggested the spending was beginning to translate into commercial returns.

"Right now we're facing an incredible uncertainty" — Owen Lamont, senior vice president at Acadian Asset Management, speaking to CNBC

Despite the severity of the recent falls, context matters. The Philadelphia semiconductor index has risen 92% over the past 12 months, even after a nearly 20% drop in the past month. Memory chip stocks collectively remain sitting on a median gain of close to 60% since late March. Some fund managers view the pullback as an opportunity. Aberdeen's investment team noted that the sell-off has brought valuations to more attractive levels, while Morningstar's Michael Field argued the decline reflects sentiment more than any deterioration in underlying business performance. The central question for markets is no longer whether AI spending will be large — it will be extraordinary — but whether it will be repaid, and how soon.

This article is free to read. It always will be — no paywall, no account, no tracking.