Two snapshots, one morning\n\nThe first version of Monday's story was blunt: Asian share markets fell as unease over the enormous cost of the AI build-out undermined confidence, Reuters reported — the Nikkei down 2%, South Korea down 1.6%, the MSCI Asia-Pacific index outside Japan down 0.3%.\n\nReuters then updated its own wrap. In the later snapshot, the Nikkei was down 0.7%. South Korea was not down at all; its benchmark had added 0.6%. Taiwan was down 1.1% and the MSCI Asia-Pacific index outside Japan was 0.2% lower. Wall Street futures, positive in both versions, were firmer in the second: Nasdaq futures up 1.3%, against 0.8% earlier.\n\nNeither snapshot is wrong. They were taken at different times in a live session, and a market that opens weak and steadies will produce both. The timestamps attached to them — first published 00:38 UTC, updated 05:21 UTC — are the wrap's publication and update times, not necessarily the moment each index level was recorded. The risk sits in how live figures travel: an accurate early snapshot can become outdated within hours, and the two versions establish only that the reported figures changed. They do not establish which version circulated more widely, or that readers were systematically misled. Anyone meeting these numbers in a syndicated copy or a screenshot needs the timestamp before treating them as the latest market position.\n\nOne boundary applies to everything above: both readings are intraday. This article compares the two snapshots as Reuters published them on Monday 19 October 2026; it is not a report of how the session ended.\n\n## What last week actually did\n\nStrip out Monday entirely and the underlying move is still substantial. The PHLX Semiconductor Index shed 11.4% last week, in what Reuters describes as its longest losing streak since 2008. The Nikkei lost 5.4% over the same week. South Korea shed more than 14%. The Nikkei and South Korea's benchmark are broad national indices, not AI indices; the PHLX is the more direct measure of chip-sector weakness. Those are full-week figures, settled in a way Monday's intraday readings are not, and no intraday narrowing on Monday changes them.\n\nWhat explains the week is less settled. Reuters' reporting highlighted concerns voiced by AMP economist My Bui: that AI revenues may not be sufficient to recoup the massive upfront investment, and that AI infrastructure may need to be written down as it ages and depreciates, lowering book value in the meantime. Alongside that, Bui pointed to rising government bond yields and wider corporate credit spreads increasing financing costs, making it harder to borrow, refinance and keep the build-out going. These are plausible contributors to the repricing, not a proven single cause — and the worry about whether AI revenue justifies the investment is itself partly a question about future demand and returns, not a story separable from financing. What can be said is that a higher cost of capital reprices every project that has not been built yet, a slower-moving pressure than a chip-order slowdown: orders can recover in a quarter.\n\n## The $1.2 trillion question arrives this week\n\nThe reason Monday's wobble matters is the calendar. Meta is one of five AI hyperscalers whose combined capital spending is expected to reach about $1.2 trillion next year, according to a Goldman Sachs estimate cited by Reuters — a projected combined total for the group, not an additional $1.2 trillion on top of this year's spending, and a figure for the five companies' overall capital expenditure rather than a sum broken out as AI-only. Meta, which reports on Wednesday, expects to spend $130 billion to $145 billion on capex in 2026, and Reuters reports that any increase in that guide could unsettle investors further.\n\nWednesday also brings Lam Research and Texas Instruments — two semiconductor suppliers whose results speak directly to whether the build-out is still accelerating — and Tesla. Results from SpaceX are due Tuesday, per Reuters. Overall S&P 500 earnings are still expected to rise 30%, largely driven by technology companies, which is precisely the tension: the profits are real and large, and the spending plan set against them is larger. Monday's later intraday reading in Seoul does not settle that argument. Wednesday's numbers are the first evidence either way.\n\nThe same morning carried a counterweight, noted here without any causal claim attached: China's economy grew 4.5% in the third quarter, just above forecasts, with industrial output growth at 5.5%, per Reuters. The morning was not uniformly negative, and nothing in the reporting ties the Chinese data to the steadier later readings.

Sources

  1. [1] Reuters (Wayne Cole), "Asian stocks slip, Wall St and bonds benefit from oil dip", Sydney, 19 October 2026, first published 00:38 UTC and updated 05:21 UTC — updated-snapshot full text: Nikkei -0.7%, South Korea +0.6%, Taiwan -1.1%, MSCI Asia-Pacific ex-Japan -0.2%, PHLX Semiconductor Index -11.4% last week, the My Bui comments, SpaceX results due Tuesday, Meta / Lam Research / Texas Instruments / Tesla reporting Wednesday, the Goldman Sachs $1.2 trillion estimate, Meta's $130–145 billion 2026 capex, and China Q3 GDP +4.5% with industrial output growth 5.5%Read source
  2. [2] Reuters, early-snapshot version of the same wrap, via BusinessToday Malaysia, "AI Spending Fears Send Asian Stocks Lower, Oil Eases", 19 October 2026 — early snapshot: Nikkei -2.0%, South Korea -1.6%, MSCI Asia-Pacific ex-Japan -0.3%, and the S&P 500 scope for the expected 30% earnings riseRead source