Analyse fra Goldman Sachs, bearbejdet til dansk resume
Analysen beskriver et marked, der stadig tror på den langsigtede AI-fortælling, men hvor selve markedsstrukturen er blevet væsentligt mere sårbar. Hovedpointen er ikke, at AI-temaet nødvendigvis er brudt sammen fundamentalt, men at positionering, crowding, korrelationer og makroforhold tilsammen har skabt et miljø, hvor selv en relativt begrænset udløsende faktor kan føre til meget voldsomme bevægelser.
Fredagens kursfald bliver fremstillet som den første egentlige stresstest af AI-handlen. Det skete efter en periode med ekstrem styrke i især teknologi- og semilederaktier. Markedet var på flere målestokke usædvanligt strakt: SOX lå meget langt over sit 200-dages glidende gennemsnit, Nasdaq havde sin værste dag siden april, og S&P 500 var steget i et tempo, der historisk ligger helt i top. Det betyder, at markedet ikke blot var optimistisk – det var teknisk og positioneringsmæssigt ekstremt udspændt.
Samtidig havde optionsmarkedet udviklet sig til et næsten euforisk setup. Normalt betaler investorer mere for beskyttelse mod fald end for upside, men her var forholdet i praksis vendt om i gennemsnittet for S&P-aktier: calls blev handlet dyrere end puts. Det tolkes som et klart tegn på aggressiv jagt på mere optur, især drevet af AI-momentum. Flows og prime broker-data bekræftede samme billede: hedgefonde og andre investorer havde i ugerne op til faldet fortsat købt aktier massivt, især i USA og Asien, mens risk appetite-indikatorer var tilbage på niveauer, man ikke havde set siden 2021.
Et særligt vigtigt advarselssignal i teksten er, at crowdingen ikke bare er høj – den er synkroniseret på tværs af strategier og regioner. Fundamentale og systematiske long/short-managere ligger usædvanligt tæt i deres afkastmønstre, og det samme gælder amerikanske og asiatiske fonde. Det betyder, at porteføljer, der på papiret burde være diversificerede, i praksis i stigende grad er eksponeret mod de samme temaer – især AI. Når korrelationerne stiger på den måde, øges risikoen for samtidige tvungne nedbringelser af risiko, hvilket kan forværre et fald kraftigt.
På selve nedtursdagen begyndte investorerne at købe mere beskyttelse, særligt i meget kortløbet SPX-downside og VIX-upside. Put-volumen steg markant, og frygtmålinger rykkede tydeligt op. Men forfatteren vurderer, at det ikke lignede fuld kapitulation. Der kom altså en reaktion i form af hedging, men ikke i et omfang, der svarer til klassiske panikfaser. Det er vigtigt, fordi det antyder, at markedet endnu ikke er renset fuldt ud for optimisme eller risikovillighed.
Samtidig understreger analysen, at dette ikke nødvendigvis er en gentagelse af 2000-boblen. Den fundamentale baggrund er anderledes stærk. S&P 500’s forventede indtjening er steget markant, og indtjeningsrevisionerne har faktisk været stærkere end indeksstigningen. AI-investeringer, kapacitetsbegrænsninger, datacenterudbygning, energibehov og halvlederefterspørgsel bliver fremhævet som reelle, underliggende drivkræfter. Med andre ord: den strukturelle AI-case er stadig intakt.
Det svage punkt ligger i stedet i makro og markedsteknik. Analysen peger på et mere udfordrende regime med higher-for-longer-renter, stigende obligationsudstedelse, strammere likviditet og en mere høgeagtig Fed-forventning efter stærke nøgletal. Indtil nu har selskabsspecifikke og indtjeningsdrevne medvinde overskygget makromodvinden, men fredagens bevægelse viste, at renter igen betyder noget for aktiemarkedet. Det gør kommende inflationstal og andre makrobegivenheder ekstra vigtige.
Et andet interessant punkt er, at den såkaldte broadening trade faktisk virkede under overfladen. Aktier uden for AI-segmentet holdt sig relativt bedre, og visse ex-AI-indeks klarede sig endda positivt på dagen. Men analysen tolker ikke dette som et ægte udtryk for ny begejstring for cykliske aktier eller bred økonomisk optimisme. Snarere tyder det på, at investorer lukkede nogle meget ensidige positioner: long AI, short resten. Det gør bevægelsen mere til en faktorrotation eller unwind end til et klassisk, bredt markedssalg.
Til sidst samles det hele i en vurdering af, at markedet er blevet mere skrøbeligt. Kombinationen af ekstrem momentum-crowding, høj korrelation, detailleverage, smal markedsbredde, ændret optionsstruktur og mere udfordrende makroforhold skaber en situation, hvor fremtidige bevægelser kan blive langt mere voldelige, end indeksniveauerne alene antyder. Derfor er budskabet ikke, at AI-historien er slut, men at risikoen for skarpe og hurtige factor unwinds er steget betydeligt.
De fem vigtigste konklusioner
1. AI-handlen blev endelig stresstestet
Fredagens fald var den første reelle test af, hvor robust AI-drevne markedspositioner faktisk er. Det skete efter en ekstrem optur, hvor især tech og semis var blevet historisk overkøbte. Pointen er, at markedet ikke faldt fra et neutralt udgangspunkt, men fra et meget sårbart og overstrakt niveau.
2. Den største risiko er crowding og korrelation – ikke at AI-temaet er dødt
Analysen siger klart, at kernerisikoen ikke er et fundamentalt kollaps i AI-casen. Risikoen er, at for mange investorer globalt ligger i de samme handler på samme tid. Når både systematiske og fundamentale fonde, på tværs af regioner, i praksis ejer de samme temaer, kan små chok udvikle sig til store synkroniserede drawdowns.
3. Markedet begyndte at hedge, men har sandsynligvis ikke kapituleret
Der kom tydelig efterspørgsel efter downside-beskyttelse fredag, og put-volumener eksploderede. Alligevel var panikniveauet ifølge analysen ikke oppe i zonen for ægte stressbegivenheder. Det antyder, at markedsdeltagerne er blevet mere nervøse, men ikke at de allerede har renset risikoen helt ud af systemet.
4. Makroforholdene er blevet mere negative
Higher-for-longer-renter, strammere likviditet og mere høgeagtige forventninger til centralbanken gør miljøet mindre støttende for risikable aktiver. Selvom indtjening og selskabsspecifikke forhold stadig er stærke, er markedet nu mere følsomt over for inflationstal, renter og finansielle forhold end tidligere i rallyet.
5. AI-fortællingen er fundamentalt intakt, men markedet er meget mere skrøbeligt
Indtjeningsrevisioner, capex, halvlederefterspørgsel og infrastrukturelle flaskehalse peger stadig på, at AI-temaet har reel økonomisk substans. Men selv en stærk strukturel fortælling kan give store kursfald, hvis positioneringen er for ensidig. Derfor er konklusionen, at AI ikke nødvendigvis er “færdig” – men at vejen frem kan blive langt mere volatil og brutal.
Kort samlet vurdering
Teksten argumenterer for, at vi er gået fra et marked præget af eufori og ensidig positionering til et marked, hvor den samme optimisme nu gør prisdannelsen ustabil. Det mest sandsynlige problem er ikke et totalt opgør med AI-temaet, men en periode med skarpere rotationer, hurtigere drawdowns og mere voldelige bevægelser, især hvis makrodata fortsætter med at presse renteudsigterne op.
———————————————————————————————————–
Uddrag fra Goldman
Following top Goldman Sachs quant trader, Brian Garrett’s, post-mortem must-read yesterday, Goldman’s flows guru, Lee Coppersmith, warns of a considerably more “fragile” market amid conditions “materially more violent” moves ahead.
Some big takeaways from this past week:
- The market finally stress-tested the AI trade.
- The stress-test happened from one of the most extended positioning / momentum setups since the Dot-Com era.
- We saw demand for hedges Friday afternoon and put volumes exploded, but hard to call it full capitulation.
- The “broadening” trade quietly worked underneath the surface, though likely more as an AI-long / ex-AI-short unwind than true cyclical enthusiasm.
- The Macro is becoming less supportive (higher-for-longer rates, rising issuance, tighter liquidity backdrop), but micro/earnings still matter more for now.
- The real risk here is not “AI is dead.” It’s that correlations and crowding across global equity books have become dangerously synchronized.
Friday marked the worst 1-day move for the NDX since April ’25 and came against one of the most technically extended setups we have seen in years. As recently as Wednesday’s close, SOX was trading +76% above its 200dma – the highest reading since March 2000. Friday’s selloff compressed the stretch by ~22 percentage points in a single session.
At the index level, SPX had rallied +15% in two months prior to Friday – a 99th percentile move relative to history since 1980. Relative to realized vol, the rally was the strongest in 50+ years with a return/vol ratio approaching 4.
The options market backdrop had also become outright euphoric. Prior to Friday, S&P average single-stock 1m put/call skew had actually inverted. Calls were trading over puts. Said differently, investors were paying more for upside than downside across the average S&P single stock – an unprecedented setup and one of the clearest signs yet of how aggressively upside chasing / AI momentum had taken over the tape.
Flows and positioning data told a very similar story: GS PB data showed global equities saw the largest net buying in nearly 4 months (+1.5 SD vs 1yr avg) from May 29th through June 4th. All major regions were net bought, led by North America and EM Asia. Hedge funds net bought US equities for a third straight week and at the fastest pace in over 6 months (+1.0 SD), while single stocks saw the largest net buying in 3 months (+1.1 SD).
At the same time, GS Risk Appetite Indicator moved back above 1.2 – the highest level since 2021. Very high RAI levels are not necessarily a bearish signal, but average returns and hit ratios for positive returns are somewhat lower, especially on a 1-3 month horizon, and risk of corrections tends to pick up.
The crowding increasingly looks synchronized across books – our PB risk team highlighted that realized 3m alpha correlations between Fundamental and Systematic L/S managers are now in the top decile of the past decade. Correlations between US and Asia Fundamental managers have also reached all-time highs, largely driven by the AI trade. Elevated cross-strategy correlations materially increase the probability of synchronized drawdowns across books that are otherwise assumed to be diversified.
Momentum Winners are now more long crowded than at any point on record while the short side remains remarkably under-owned — an unusually asymmetric setup that leaves the factor vulnerable to sharper unwinds if positioning starts to reverse (h/t GS’ Marco Laicini).
Flows-wise, we did start to see demand for protection emerge as the tape deteriorated Friday afternoon, particularly in very short-dated SPX downside and VIX upside. Put volumes exploded into the close. That said, the magnitude of the hedging flow still felt relatively modest versus the size of the underlying move lower. Our Vol Panic Index jumped from 1.6 on Thursday to 6.5 on Friday. That is a real move higher in fear/panic metrics, but still well below the 8-10 panic zone we saw during prior true stress events.
The repricing in options was notable though. After Friday realized ~5x the daily SPX straddle to the downside, Monday’s implied move reset to ~1.1% while Friday’s straddle moved out to ~2.1%, reflecting a meaningful repricing of near-term event risk into a very busy macro week ahead.
S&P implied volatility was up in short-dated tenors, but down in August and beyond
The fundamental backdrop also remains materially different from 2000. Consensus forward S&P EPS estimates have risen +16% YTD, actually outpacing the +8% move in the index itself. Unlike prior speculative peaks, the rally has largely been driven by earnings revisions rather than pure multiple expansion. AI capex, memory tightness, datacenter bottlenecks, power constraints, and semiconductor supply/demand dynamics all still look very real.
The macro backdrop around the trade is becoming less supportive though. Following the stronger NFP print, our economists now expect no Fed cuts this year and pushed the final two cuts in their baseline forecast out to June and December ’27. GS now assigns a 20% probability to modest hikes, up from 10% prior, and sees 2yr / 10yr yields ending 2026 at 3.8% and 4.4%, respectively. The market is now being forced to consider a world where growth remains firm enough and inflation sticky enough to keep policy restrictive for materially longer. Up until now, micro tailwinds had almost completely overwhelmed macro headwinds. This week was one of the first reminders that rates still matter again.
Following Friday’s NFP-driven repricing across the rates curve, this week’s CPI (Wednesday) and PPI (Thursday) take on more importance as the market debates whether the stronger growth / stickier inflation backdrop continues to push the Fed distribution more hawkishly.
We’re also watching the highly anticipated IPO, which could become another important liquidity / positioning event given the increasingly active issuance backdrop.
Beneath the surface, the “broadening” trade quietly worked. SPXXAI, the S&P 500 ex-AI index, actually finished Friday UP while SPX fell -2.6% and RSP declined -1.4%. Small Caps ex-AI also materially outperformed broader RTY. To me, that likely says less about investors suddenly loving cyclicals and more about how crowded the AI-long / ex-AI-short positioning backdrop had become. Friday looked much more like a violent factor unwind than indiscriminate equity selling.
Reverse dispersion conversations also continue to accelerate. Index performance remains relatively contained while single-stock, factor, and thematic volatility underneath the surface continues to expand. Crowding in AI leadership alongside elevated positioning and correlations increasingly looks like an unstable equilibrium.
The secular AI story remains intact. Earnings revisions remain strong. Flows had still been aggressively supportive coming into the week. Friday finally produced a real volatility response, but not enough to suggest a fully hedged market backdrop. In Korea, using historical earnings cyclicality and valuation, KOSPI cyclical downside would be roughly where the market is trading, suggesting risk/reward remains positively skewed.
But the setup is clearly more fragile.
The combination of momentum crowding, elevated correlations, higher-for-longer rates, growing issuance expectations, retail leverage, recently inverted single-stock skew, and increasingly narrow breadth all create an environment where factor unwinds can become materially more violent than index-level volatility implies.
2 md. adgang for
2 x 49 kr.
Få straks adgang til denne artikel og derefter 2 måneder til alle artikler på ugebrev.dk
- Alle artikler på ugebrev.dk
- Om investering, finans, ledelse, samfundsansvar, life science og Bestyrelsesguiden.dk
- Daglige nyhedsmails med nyheder og analyser
Tilbuddet gælder til 31. juni 2026. Abonnement fortsætter til normalpris på 249 kr. efter bindingsperiode på to måneder. Opsig når du vil - til udgang af den anden måned. Tilbud gælder kun, hvis du ikke har haft abonnement på ØU udgivelser de seneste tre måneder
Allerede abonnent? Log ind her
