What four sessions did
The American semiconductor index SOX fell as much as 6 percent on Tuesday, a fourth straight losing session, and the Nasdaq-100 ended 9.7 percent below its record, just short of the threshold that counts as a correction. South Korea's Kospi closed down nearly 11 percent and triggered its eighth circuit breaker of 2026. These are observed price moves, and on their own they do not say which piece of information is being priced.[1]
Two supply-side series published in the same week look the other way. Third-quarter DRAM contracts are settling 20 percent to 30 percent higher this month. Alphabet raised its capital-expenditure guidance for this year to 205 billion dollars, from 91 billion dollars in 2025. Moody's expects the six largest cloud-infrastructure companies to spend roughly 785 billion dollars this year.[1]
What the operating number says
The one large operating release that followed pointed the same way. SK hynix reported second-quarter revenue of 79,318.7 billion won and operating profit of 60,542.6 billion won, an operating margin of 76 percent, with revenue up 51 percent and operating profit up 61 percent on the prior quarter. The company also said mass shipments of HBM4 had begun and that it had signed long-term agreements with around 10 customers.[2]
Contract pricing, capital-expenditure guidance and the quarterly operating margin all fail to show near-term cash inputs deteriorating. On that reading, four sessions of losses fit an unwind of crowded positioning better than a revision to expected demand. A second reading is equally consistent with the same data: the market may already be discounting a digestion phase in 2027, and such an expectation would not yet appear in current contracts.[1], [2]
The observation that separates them
Gil Luria, a technology analyst at D.A. Davidson, said there was a great deal of panic around the artificial-intelligence investment and that the panic appeared indiscriminate. That is interpretation, and it carries a testable direction: an indiscriminate selloff carries names whose input economics are improving down alongside those whose are not.[1]
The distinguishing observation sits at the contract table. If fourth-quarter DRAM contracts settle flat or lower and the cloud-infrastructure companies trim capital-expenditure guidance at the next reporting round, the decline is pricing a revised set of fundamentals. If neither condition arrives by the end of November and the spending plans hold, four sessions of losses remain a positioning move.[1]