Macro
The US growth-inflation mix deteriorated at the margin, although the apparent stickiness in PCE may overstate underlying inflation pressure. Core PCE rose to 3.34% YoY and headline PCE to 3.70%, while real consumption was flat, retail sales fell 0.6% MoM and Chicago PMI dropped to 47.1. However, core PCE is now 88bps above core CPI, largely because PCE contains a small number of unusually large price increases: the gap between the two measures falls to only 29bps on a trimmed-mean basis. If that positive PCE skew normalizes over the next 6-12 months, underlying inflation could fall faster than the headline PCE rate currently suggests. Labor remains the main counterweight to the slowdown, with initial claims at just 203k, keeping the picture closer to slower growth than recession.
Jackson Hole reinforced the Fed’s inflation reaction function but did not fundamentally change the macro outlook. In his first Jackson Hole speech as Chair, Kevin Warsh reaffirmed the 2% PCE target, said financial conditions do not appear meaningfully restrictive and left open further tightening if inflation fails to improve, while deliberately avoiding guidance for September. The important macro point is therefore not the immediate market repricing but that policy remains highly dependent on whether the recent inflation persistence proves genuine or, as the PCE/CPI composition suggests, partly temporary. The resulting rates and market-pricing implications are covered separately in the RATES update.
The US consumer is increasingly bifurcated, with aggregate resilience masking greater pressure at the lower end. Real disposable income rose 0.4% MoM and three-month annualized real spending remains 3.3%, but the savings rate is only 3.0%, card-spending growth slowed to 5.0% from 6.3%, and consumer-credit stress remains elevated. Gasoline around $3.99/gallon, up roughly 47% YTD, adds another regressive burden, leaving premium consumption substantially better supported than broad consumer cyclicality.
Oil’s selloff reflects lower geopolitical risk pricing, not normalization of physical supply. Brent fell 5.4% to $89.31/bbl and WTI 4.2% to $83.40 following the Iran-Oman corridor announcement, yet regional exports remain only around 9-11mb/d, roughly half pre-war levels, with 2.5-3mb/d of Gulf refining capacity still offline. More structurally, Hormuz is accelerating investment in alternative pipelines, transport corridors, inventories and supply-chain redundancy. That capex is inflationary in the near term, but the eventual creation of redundant capacity could become disinflationary, making the geopolitical shock simultaneously a near-term inflation risk and a longer-term supply response.
Europe’s recovery is broadening, but its energy exposure limits how aggressively investors should extrapolate the improvement. Eurozone composite PMI reached 52.1, Germany’s ifo rose to 88.8, German Q2 GDP was revised to +0.3% QoQ, and French consumer confidence improved for a third month. However, Spanish inflation accelerated to 4.5% YoY, EU gas storage remains 11.1pp below last year, and Gulf suppliers account for 31.1% of EU refined-product imports, leaving Europe disproportionately vulnerable to another Middle East supply shock.
China remains an export and technology growth story rather than a broad domestic-demand recovery. Exports surged 27% YoY and high-tech industrial profits are up more than 50% YTD, but retail sales grew only 0.6%, fixed investment fell 6.7%, manufacturing PMI slipped to 49.2, and new RMB lending contracted by RMB340bn. The investable growth impulse therefore remains concentrated in AI, advanced manufacturing and exporters rather than broad Chinese domestic demand.
Emerging Asia is benefiting from the AI and semiconductor cycle, but this is not yet a synchronized global demand recovery. South Korean exports surged 62% YoY in the first 20 days of August, reinforcing the strength of the regional technology supply chain, while Citi expects Emerging Asia to grow 5.1% in 2026. Global resilience is therefore increasingly being driven by capex, technology and exports rather than households, property and traditional domestic demand.
Trade and geopolitical fragmentation remain persistent supply-side risks rather than isolated shocks. US tariffs already cover roughly $20bn of Canadian exports, equivalent to around 0.9% of Canadian GDP, while the proposed 50% tariff on Canadian autos and parts from January 2027 could significantly disrupt integrated North American production if implemented. More broadly, the Iran conflict reinforces an existing shift toward supply-chain redundancy, defense and infrastructure investment across a more multipolar world, supporting capex but making the global economy structurally more expensive to operate in the near term.
The dominant regime remains slowing but positive growth, elevated supply-side risk and inflation that looks sticky today but may be closer to an inflection point than headline PCE suggests. That distinction matters: the economy is not yet recessionary, while the 88bp PCE-CPI gap argues against treating every high PCE print as evidence of renewed underlying inflation. At the same time, geopolitical fragmentation, energy disruption and trade barriers continue to generate new inflation shocks, leaving investors with a regime dominated by dispersion rather than a clean Goldilocks or stagflation call.
Rates
For investors, the defining feature of the week was a policy-driven flattening of the Treasury curve rather than a conventional directional rates move. Kevin Warsh repriced the front end through a hawkish Jackson Hole speech, while Scott Bessent’s expanded long-end buyback program compressed term premium at the back end. The 2Y yield rose 11 bps and the 5Y 5.5 bps, while the 10Y fell 1.4 bps and the 30Y declined 6.5 bps. The move was also distinctly US-specific: Bund, Gilt and JGB yields all increased, reinforcing that Treasury outperformance reflected domestic policy and technicals rather than a global duration rally.
Warsh’s Jackson Hole message materially raised the risk that front-end policy remains too easy and that further tightening may be required. In his first Jackson Hole address as Fed Chair, he reaffirmed the 2% PCE objective as a firm, fixed target, argued that inflation had not slowed sufficiently and said financial conditions were difficult to describe as restrictive. He also proposed replacing conventional forward guidance with broader governing principles and deliberately avoided a mechanical reaction function, preserving greater policy flexibility. Markets interpreted this as materially more hawkish than his July press conference: on Friday the 2Y rose 11.4 bps, versus 4.4 bps in the 10Y and only 1.3 bps in the 30Y, compressing 2s10s by roughly 7 bps. The reaction suggests investors increased near-term tightening risk while simultaneously viewing Warsh’s inflation stance as supportive of long-run credibility.
September is now genuinely live, with the implied probability of a September hike rose from 39.6% to 59.8%, while the December implied policy rate increased 11.9 bps to 4.013%. By year-end, markets are pricing roughly 38 bps of cumulative tightening, versus approximately 26 bps a week earlier. Cuts were not priced entering the week, so the adjustment was a straightforward addition of further tightening into the curve.
At the long end, Bessent has created a new technical support mechanism for Treasuries without improving the underlying fiscal fundamentals. Treasury plans buybacks of at least $4 billion per operation in 10Y, 20Y and 30Y nominal coupons between September 9 and November 4, at least doubling the prior cap. Funding may come through additional bill issuance or potentially the Treasury General Account, where estimates suggest $80–200 billion of usable capacity, while Treasury has already indicated that long-term issuance will not increase for at least several quarters.
More importantly, Bessent is effectively shortening the maturity of the government’s balance sheet in an attempt to influence long-term yields. Unlike traditional, broadly duration-neutral liquidity buybacks, removing long bonds while funding purchases with bills or cash replaces duration risk with refinancing risk and amounts to a macro wager that future rates will be lower. If correct, Treasury can refinance more cheaply later; if inflation remains sticky and the Fed tightens, it will instead roll more short-dated debt at higher rates, with roughly one-third of marketable federal debt already maturing within 12 months. Critics including Stanley Druckenmiller argue that elevated yields reflect fundamentals rather than market dysfunction: inflation remains persistent, the deficit is near 6% of GDP, federal debt exceeds $40 trillion, and annual interest expense is approaching $1.1 trillion. The irony is that Bessent previously criticized Janet Yellen for relying excessively on bills and politicizing issuance, while the current strategy represents an even more explicit maturity-shortening intervention.
Market pricing supports the view that the long-end rally was a technical term-premium move rather than an inflation-expectations rally. The 10Y real yield was almost unchanged at +0.2 bps, while the 10Y breakeven declined only 1.5 bps. By contrast, the 5Y real yield rose 7.6 bps, consistent with Warsh’s hawkish message being concentrated in the front end and belly. The 30Y decline therefore looks primarily like real term-premium compression from reduced effective duration supply. Positioning amplified the move: speculative investors entered the week with aggressively larger duration shorts across long-bond, 10Y and 5Y futures, leaving the market vulnerable to a squeeze. Part of the rally had already faded by Thursday, suggesting the intervention changed positioning more quickly than fundamentals.
Treasury auction demand reinforced the same curve distinction: strong absorption at the front end, but persistent weakness in the belly. The $69 billion 2Y auction stopped 0.4 bps through the when-issued level, and bill auctions were robustly absorbed, supporting Treasury’s assumption that additional short-end issuance can be financed without immediate disruption. By contrast, the $70 billion 5Y auction tailed by 0.2 bps, marking a tenth consecutive weak result and reinforcing persistent softness in intermediate maturities.
For portfolio construction, the strongest expression remains curve- and maturity-specific rather than an outright duration call. The front end now embeds materially more Fed tightening, though there is still a credible case that markets are pricing more hikes than will ultimately be delivered. The long end benefits from direct Treasury technical support and attractive starting yields, but the recent rally appears more technical than fundamental and remains vulnerable if fiscal concerns reassert themselves.
The broader institutional shift may ultimately matter more than this week’s yield moves: Warsh is reducing the Fed’s reliance on forward guidance just as Bessent becomes more willing to influence long-term yields through debt management. Week 35 therefore produced a clear division of influence across the curve: Warsh owned the front end; Bessent owned the long end. The key question is whether Treasury can continue suppressing term premium while the Fed remains focused on inflation and potentially raises the refinancing cost of an increasingly short-dated government debt stock.
Credit
Credit remained resilient at the index level, but dispersion increased materially beneath tighter spreads as investors became more selective around AI financing, software refinancing and private-credit liquidity. US IG OAS tightened 3bps to 78bps and returned +0.32%, while HY tightened 9bps to 260bps and returned +0.27%, despite the late-week rates selloff. Stress remained concentrated in lower-quality credit, with CCC OAS around 1,183bps and technology-loan distress rising sharply. The weekly signal was therefore not broad risk aversion but a widening gap between strong credits that continue to access abundant capital and structurally complex or lower-quality borrowers that increasingly require higher yields, stronger documentation and greater lender protection.
Record August IG supply was absorbed without destabilizing spreads, but weakening order quality and softer HY technicals suggest September will increasingly test pricing rather than absolute market capacity. US IG issuance reached $36.0bn during the week, versus roughly $20bn expected, taking August supply to a record $190.8bn, more than 40% above the previous August record of $136bn in 2020; HY issuance was $2.85bn for the week and $37.3bn for August. September IG forecasts reach $250bn, potentially a record, while SoftBank is considering a $20bn dollar/euro transaction partly linked to its OpenAI financing. Recent new-issue concessions compressed from roughly 11bps to 5bps, but order-book attrition remained around 40-45% versus 23% YTD, indicating that headline books increasingly overstate true demand. IG funds still recorded a $906.7mn inflow, the 19th consecutive positive week, but HY saw a $1.52bn outflow, the largest since May, while leveraged loans attracted $406.5mn.
AI infrastructure is becoming a distinct credit risk premium, with investors demanding higher compensation for project, power, construction and refinancing risk rather than treating data-center debt as conventional high-grade technology exposure. Relevant spreads have widened roughly 20-70bps, several IG data-center deals have fallen 2-5 points in secondary trading, and approximately 80% of data-center securities issued since early 2025 now trade wider than their original issuance spreads. Blackstone-backed QTS Realty priced $3.9bn of IG-rated secured debt at roughly 7.23%, while BlackRock’s Meta data-center financing yielded around 7.53%; Zenith Arc’s $2.25bn financing weakened after issuance, and a roughly $1.15bn CoreWeave-linked HY deal expected in September will provide another clearing-price test. Many structures assume refinancing after roughly two years once projects are completed, but that assumption becomes increasingly fragile as investors look toward potentially $500bn of AI-related financing in 2027. With hyperscaler project returns potentially above 25%, creditors are increasingly questioning why they should accept narrow spreads while retaining much of the tail risk, pushing the market toward clearer segmentation between stabilized IG assets and speculative or unbuilt projects requiring HY-like pricing, stronger collateral or tighter structures.
Software credit is showing the clearest shift in negotiating power toward lenders as AI disruption compounds an already difficult refinancing cycle. Private-equity sponsors face more than $55bn of software debt requiring refinancing, while the share of technology loans trading below 80 cents has risen to 23%, from 8.2% a year ago. Proofpoint’s refinancing required Thoma Bravo to accept around 40 creditor-friendly amendments, including tighter additional-borrowing restrictions, more extensive disclosures and regular lender engagement. Sophos is considering a higher coupon, amortization and tighter covenants to refinance more than $2bn, Conga is extending a $548mn term loan, and Gainwell has completed a $5.8bn debt overhaul. Refinancing remains available, but increasingly through shorter extensions, higher coupons and stronger creditor protections rather than simple maturity pushes on legacy terms.
Private credit is increasingly exposing the gap between reported NAV stability and executable secondary liquidity, even as some redemption pressure begins to moderate. Non-traded BDCs face an estimated $15bn redemption backlog across more than a dozen funds, equivalent to roughly 15% of AUM, while quarterly liquidity is typically capped near 5%. Cox Capital offered to purchase up to $90mn of BDC shares at an average 26% discount to NAV, yet received less than $5mn of sell orders, suggesting investors prefer waiting in redemption queues to crystallizing steep discounts. Golub’s redemption requests eased to 4.8% from 8.5%, but non-traded BDC fundraising fell to only $2bn in Q2, down 82% YoY from $11bn. Underlying stress also remains elevated: a direct-lending shadow default rate of roughly 16% compares with around 3.7% for HY bonds and 3.3% for broadly syndicated loans, while BlackRock’s TCP Capital sold $523mn of loans, nearly half its portfolio, into a Pantheon-backed secondaries vehicle. Dedicated secondaries capital can improve liquidity, but the discounts required to clear transactions make marks, recovery assumptions and redemption structures increasingly important credit variables.
Portfolio positioning should favor credits where higher yields now compensate for identifiable and underwritable risks, while remaining cautious toward structures dependent on aggressive refinancing assumptions, weak creditor protections or opaque NAVs. Tight headline IG and HY spreads leave limited compensation for broad beta, but dispersion is creating opportunities in selected AI infrastructure transactions offering HY-like yields with stronger collateral and in software refinancings where lenders are obtaining materially better economics and documentation. The more vulnerable areas remain speculative data-center projects, distressed technology loans and private-credit structures where reported marks coexist with large secondary discounts. The structural shift is increasingly clear: market capacity remains substantial, but creditors are becoming less willing to provide it without being paid explicitly for complexity, liquidity and refinancing risk.
Equities
US equities advanced at the headline level, but deteriorating breadth showed that the rally had narrowed sharply toward mega-cap growth. The S&P 500 gained 0.50%, Nasdaq Composite 0.85% and Dow 0.55%, while the equal-weight S&P 500 fell 0.44%, creating a 94 bp cap-weight advantage, and the Russell 2000 declined 1.46%. The median S&P 500 constituent fell 0.39%, confirming that the positive index return was not representative of the underlying market. International performance was mixed: Germany’s DAX led at +1.66%, the Nikkei gained 0.64% locally but fell 0.13% in USD, the FTSE 100 was flat at +0.08%, MSCI Emerging Markets rose 0.05%, and the Hang Seng declined 1.50%. The combination of rising large-cap indices, negative equal weight and weak small caps points to concentration and liquidity preference rather than broad risk appetite.
Sector performance reinforced the same concentration signal, with Technology, Communication Services and Financials providing almost all of the positive leadership. Communication Services (+1.43%) and Information Technology (+1.30%) led on large-cap internet, software and AI exposure, while Financials gained 1.08% as the higher-for-longer rate repricing improved the relative earnings backdrop for parts of the sector. Utilities were resilient at -0.09%, suggesting AI-related power demand continued to offset some duration pressure. Weakness was broad elsewhere: Health Care fell 1.98%, Industrials 1.73%, Energy 1.51% and Real Estate 1.33%, while Consumer Discretionary (-0.69%), Materials (-0.67%) and Staples (-0.63%) also declined. Energy’s reversal after the prior week’s strength reflected the fading oil-price tailwind as geopolitical risk premium eased.
The 94 bp cap-weight/equal-weight divergence was driven by unusually strong performance across the largest index constituents, amplified by a concentrated enterprise-software earnings wave. Microsoft rose 6.3%, Meta 5.1%, Apple 3.4%, Amazon 3.0%, Berkshire Hathaway 1.9%, JPMorgan 1.7%, Nvidia 1.3% and Alphabet 0.5%, with Tesla (-3.9%) the main mega-cap exception. Software generated the most dramatic individual moves: Salesforce surged 22.4%, CrowdStrike 13.8%, ServiceNow 12.6%, Veeva 11.6%, Synopsys 11.2% and Fortinet 8.1%. Salesforce reported adjusted EPS of $5.90 versus $3.27 expected, raised FY2027 EPS guidance to $16.67-$16.71 versus $14.14 consensus, and increased revenue guidance to $46.1-$46.4B, while CrowdStrike beat Q2 expectations and lifted FY2027 revenue guidance to $5.99-$6.01B. The week’s narrow leadership was therefore supported by genuine earnings upgrades, but concentrated in a small group carrying disproportionate index weight.
Nvidia delivered the week’s most important confirmation that AI infrastructure demand remains exceptionally strong, even though its modest +1.3% weekly return concealed a much larger two-day repricing. Q2 FY2027 revenue reached $96.22B, up 106% y/y and roughly 4.2% above consensus, with Data Center revenue of $89.0B beating expectations by approximately 3.7%; adjusted EPS of $2.22 was about 5.7% ahead, while adjusted operating income of $63.96B beat by roughly 4.5%. Q3 revenue guidance of $108B was around 3% above consensus, but the decisive surprise was management’s unusually long-dated outlook for roughly 70% FY2028 revenue growth, approximately 25 percentage points above Bloomberg consensus. Purchase commitments surged to $279B from $119B sequentially, while management identified supply rather than demand as the binding constraint. Vera Rubin is already expected to account for roughly 20% of Data Center revenue in Q3, reinforcing the conclusion that the AI infrastructure cycle is extending into the next architecture rather than simply exhausting the current generation.
Nvidia’s post-earnings reversal nevertheless showed how demanding expectations, regulatory risk and higher discount rates can overwhelm even exceptional fundamentals over short horizons. Shares surged 8.7% on Thursday, closing at $227.98 and adding roughly $420-440B of market value as investors absorbed the FY2028 outlook; 24 analysts subsequently raised price targets, with no cuts. On Friday the stock fell 4.6% to $217.55 after the WSJ reported that Nvidia had paused its AI Compute Partnership revenue-sharing programme amid potential antitrust concerns, while the hawkish rate repricing simultaneously pressured high-multiple technology. The weekly +1.3% return therefore understates the event: Thursday was a fundamental rerating on stronger multi-year demand visibility, while Friday introduced a separate financing/regulatory overhang and macro valuation shock. The broader AI trade also became more differentiated: Microsoft, Meta, Amazon and Oracle outperformed as investors rewarded cloud and software monetisation, while AMD fell 1.6%, Broadcom was flat and Marvell dropped 8.6% despite beating and raising guidance after expectations had become increasingly demanding.
Breadth and factor performance confirm that this was not a conventional growth-led risk-on week, but a simultaneous preference for large, liquid winners and defensive exposure. Only 52.9% of S&P 500 constituents remained above their 50-day moving averages, versus 68.6% above their 200-day averages, indicating meaningful short-term deterioration without yet signalling a structural trend break. Low Volatility gained 0.50%, broad Value 0.40%, Growth 0.29% and Quality 0.18%, while High Beta fell 1.54% and Momentum 1.79%. The momentum drawdown is particularly important because prior winners did not move uniformly: software surged while semiconductors, industrials and other higher-beta exposures weakened. The market rewarded earnings visibility, liquidity and selected defensiveness rather than indiscriminate beta.
Investor sentiment remained calmer than the deterioration beneath the index would imply, leaving a notable disconnect between volatility and dispersion. The VIX declined from 15.13 to 14.43 despite the hawkish rate repricing, suggesting options markets did not interpret the move as an acute equity-risk event. Corporate demand also remained technically supportive as most companies entered post-Q2 open repurchase windows, although granular US equity ETF and prime-brokerage positioning data for the week remain limited. What is observable is the outcome: small caps, high beta, momentum and the median S&P 500 stock all weakened while headline volatility compressed. That combination points more toward rotation and concentration than outright institutional de-risking.
Regional dispersion remained meaningful, with Germany benefiting from improving export expectations while Hong Kong continued to struggle with technology and rate sensitivity. The DAX’s 1.66% advance was supported by the Ifo export expectations balance rising to 9.6, its highest since February 2022, continued defence spending and corporate cost discipline, including Volkswagen reducing its five-year investment plan to roughly €135B from €160B. German rate pressure was also substantially milder than in the US. The Hang Seng fell 1.50%, with Alibaba down 8.5% on August 24 and Xiaomi down 3.4% later in the week, reinforcing pressure from technology weakness and tighter financial conditions.
The portfolio implication is that headline S&P 500 resilience should not be interpreted as confirmation of a healthy broad-market advance. Earnings delivery remains strong enough to support selected software, hyperscaler and mega-cap franchises, while Nvidia’s results materially reinforce the structural AI-capex thesis. But the 94 bp cap-weight/equal-weight gap, negative median-stock return, -1.46% Russell 2000 performance and losses in momentum and high beta show that the market’s tolerance for weaker balance sheets, elevated expectations and rate-sensitive exposures has fallen. The immediate asymmetry therefore favours companies where earnings and AI monetisation continue to validate valuations while arguing for greater discipline around index concentration and indiscriminate high-beta exposure. A healthier continuation of the rally would require renewed participation from equal weight, cyclicals and smaller companies; without that broadening, index upside remains increasingly dependent on a relatively small group of highly weighted earnings winners.