Deluard's stagflation case and China's AI break

2026-10-28 · United States · Deluard's stagflation and bond case, China's AI bubble and memory, ECB, the krona and Riksbank

Daily show11:46Download MP3About the show

How it is made. Aavistus Daily is written and voiced by our own systems, every morning, from data we gather ourselves. The voices are synthetic. Every fact is traced to a source on this page, and every dated call is settled in public on the calibration page. The presenters, Nora and Callum, are named characters, not people.

Aavistus is a new framework, still under development. We will make mistakes, and every dated call is settled here in public so you can see them; the aim is to keep calibrating until the record beats the average forecaster, not to sound right.

A StoneX strategist calls a US bear market and favours long Treasuries, and Finance Bureau says China's AI bubble burst; we test both, then read the krona into the ECB.

Aavistus briefs. Independent market intelligence — satellites, flows, filings. Register free to follow new briefs.

Testing Deluard's stagflation call

Vincent Deluard of StoneX argued on Monetary Matters with Jack Farley that US inflation settles at 3.5–4%, that equities are heading into a bear market, and that long Treasuries are a better tactical place to be than stocks.

Sticky inflation we do not dispute. The "stag" half is another matter: unemployment is 4.2% (from 4.1%), initial claims fell again in the latest week, and the Sahm indicator reads zero. Credit is not confirming a bear market either: high-yield spreads widened seven basis points in the week to 23 October and are drifting toward 400, not through it.

Grind, not break, until the spread crosses 400 and a top-four hyperscaler cuts 2027 capex guidance by more than 10%. The read is wrong if the spread jumps through 400 without any capex cut. And capex is not being cut: Finance Bureau reports Microsoft offering buyouts to about a quarter of its workforce while still spending heavily on AI data centres and chips.

The long end

Our standing view is that the US 30-year grinds toward 5.75% by mid-2027. The curve gives Deluard something: over the past month the 10-year fell to 4.99% from 5.24%, the 30-year to 5.3% from 5.56%, and the 2-year to 4.7% from 4.92%. But Bloomberg reports US-friendly nations cut Treasury purchases by 60%, led by Asia — the foreign buyer stepping back is term premium rebuilding.

Position: 5.75% on the 30-year is the level. Confirmation is the 30-year back above 5.56%, where it stood a month ago, with foreign buying still cut; if the 30-year keeps falling from 5.3%, the long-bond trade is winning and the grind is delayed.

Warsh's silence

Finance Bureau reports that Kevin Warsh is declining to give forward guidance or submit a dot plot projection, and that the Fed is reviewing whether to scrap the dot plot. We agree the decision day gets jumpier; we disagree that silence is the bigger risk. The meeting is on 4 November and payrolls follow on 6 November. Our view is an upper bound at 4.75% or higher by mid-2027; it holds while unemployment stays near 4.2% and the Sahm indicator at zero, and is wrong if unemployment climbs and the Sahm indicator rises clearly off zero.

China's AI break does not travel — yet

Finance Bureau says China's STAR 50 is down more than 45% in under four months and warns of spillover into US tech.

Walking our AI-capex-to-memory chain: spending is accelerating; a record quarter at components maker Samsung Electro-Mechanics is consistent with the margin link but is not memory-margin proof; concentration remains the standing read. The link the spillover claim needs — a weekly fall of 10% or more in Samsung, SK Hynix or Kioxia — has not come, and credit has not crossed 400. The STAR 50 fall reads as domestic: Bloomberg reports state intervention widening the mainland premium over Hong Kong.

We agree the bubble has burst and disagree on transmission. For a Nordic book the exposure runs through Europe: our view is a Euro Stoxx 50 down-leg from about 6,165 by mid-2027, triggered by a memory reset, and wrong if a reset comes and the index holds above 6,100. On Chinese internet, the case for stepping in is not there until the domestic regulatory signal clears and consumer demand turns. Spillover is confirmed if Samsung and SK Hynix fall 10% in a week alongside the STAR 50; until then the China break stays in China.

The ECB, the krona and the Riksbank

If the Governing Council weighs core over headline inflation, our view is a hold at 2.5% tomorrow: headline inflation hit 3.8% in September but core only 2.5%, and German core slipped to 2.5%. The Council hiked on 16 September and does not move back-to-back without core confirming. Our view of at least 3.25% by mid-2027 stands.

The Riksbank sits at 1.75% with its own path rising to about 2.4% by early 2028. Swedish inflation was 0.3% in August, core 1.2%, and SE3 power averaging about €12/MWh this week against €71 last week pulls the next energy reading down. We do not see inflation letting the bank follow its path; the read breaks if core heads toward 2%.

The ECB at 2.5% against the Riksbank at 1.75% keeps the krona on the back foot: EUR/SEK stays above 11.00, a weekly close above 11.25 confirms, a weekly close below 11 on an ECB move falsifies. Norges Bank's policy rate is 4.5%, with Norwegian inflation at 3.2% in August and core at 3.5%; EUR/NOK stays under 11.00, and a weekly close back above 11 would break that read.

What we watch next

Whether tomorrow's ECB statement leans on core or headline inflation; the largest hyperscalers' 2027 capex guidance; the euro-area growth flash on 30 October; and US payrolls on 6 November.

Sources