Research — 27 August 2026

Verdict

The Treasury buyback escalation and Bessent's signaling have effectively removed the Fed's option of letting long-end yields tighten financial conditions; expect a steeper curve, continued hard-asset bid, and a Warsh speech that tries to reassert hawkish credibility without moving the long end.

Analysis

The paper-led thesis is that US fiscal-monetary coordination has crossed a threshold. TS Lombard argues Bessent's buyback scheme's only lasting effect is to 'cadence the march to higher yields' — it is not monetization, but it removes Warsh's choice of letting the long end do the tightening, leaving 'a steeper yield curve and a wobbling equity market' on tap; the structural driver is too much Federal debt, too little real growth, and a need for foreign capital near 3% of GDP. Bear Traps quantifies the market reaction: a $14bn buyback increase fueled a $1.5 trillion rally in bitcoin and precious metals, 30-year yields dropped 10bp, and CTAs learned Bessent will orchestrate short squeezes — a hard-asset driver rooted in late-cycle US fiscal behavior. CACIB frames the 'Bessent twist' as sparking concern the government is lowering rates through unorthodox coordination rather than tackling fiscal inflation, with the answer mattering most for the USD. CACIB's fiscal-dominance primer supplies the regime framework: fiscal dominance breaks the bond-equity hedge as term premium reprices, while financial repression favors equities but exits abruptly. The counterargument is real: GS rates research (via the FX trader call) sees the buyback impact as 'more micro than macro' — long-end yields reflect macro factors, not a point-of-curve supply imbalance, and anything substantive on the fiscal side requires legislative appetite. ING adds the market-structure risk: with markets still hawkish, the FOMC may be drawn into a hike by market pricing to avoid bond volatility. Portfolio consequence: duration is now a policy-contested risk factor; hard assets and curve steepeners are the natural expressions, with Warsh's speech the immediate event risk.

Tension

Is this monetization and regime change, or a supply-management operation that leaves the macro backdrop intact?

What changed

Bessent doubled buyback tranches in longer-dated Treasuries two weeks after announcing the quarterly program, and hinted Treasury could buy more than announced.

Desk views

TS Lombard: the buyback cadences higher yields and strips Warsh of the option to let the long end tighten conditions; curve steepening and equity wobble remain on tap.

Evidence IDs · P001

Bear Traps: $14bn of buyback escalation produced a $1.5trn crypto/precious-metals rally and a 10bp drop in 30y yields; the US is in the later stage of the fiscal cycle — the foundational hard-asset driver.

Evidence IDs · P002

GS rates research pushes back: the buyback is micro, not macro — long-end yields reflect a range of macro factors, and meaningful fiscal change requires legislative follow-through Bessent cannot deliver alone.

Evidence IDs · P019

ING: benign PCE supports no-hike, but hawkish market pricing creates the risk the FOMC hikes to avoid bond volatility; Warsh's speech is pivotal.

Evidence IDs · P021

Investor implication

Treat long-end duration as a policy-contested position rather than a clean macro hedge; hard assets retain a fiscal-credibility bid that is not purely a rates trade.

Trade expression

Curve steepeners and hard-asset exposure (gold, BTC proxies) as the fiscal-credibility expression; respect Bessent's demonstrated willingness to squeeze one-sided short Treasury positioning.

Invalidation

A Warsh speech that reasserts monetary dominance with market-following (long-end selloff tolerated, buyback follow-through absent) would undercut the fiscal-dominance framing.

Verdict

The AI trade is fundamentally supported near-term — NVDA's print and guidance were strong, and de-risking had already run its course — but the index-level thesis now rests on AI infrastructure earnings growth (55% y/y) that is unsustainable without demonstrated ROIC, making breadth and dispersion the next phase.

Analysis

The paper-led thesis has two layers. At the single-stock level, GS research sees 'a clearer path for the stock to outperform' on three factors: upside to a 70% CY27 growth guide, 72-73% medium-term gross margin guidance putting input-cost concerns to rest, and transparency on customer financial guarantees with a commitment to return over 50% of excess cash flow; the print itself beat across the board ($96.2bn revenue vs $92.4bn consensus, 75.0% GM, $89.0bn data center). At the index level, Pasquariello's framing is the essential tension: AI-infrastructure EPS grew 55% y/y vs 14% for the rest of the index in a <2% real GDP economy, and 'sustaining 55% growth is going to be very difficult' absent widespread evidence AI capex delivers true ROIC — only 2% of S&P companies quantified AI productivity's earnings impact today despite ~2-in-3 mentioning it. Positioning context matters: Coppersmith documented the largest US equity selling since Liberation Day week, net leverage at a 1-year low of 48.3%, and the largest Info Tech de-grossing in over two years — meaning NVDA was 'effectively the next major test of the market's growth narrative,' which it passed. JPM's thematic work points to the consequence: tech broadening with dispersion (software no longer trading as a monolithic AI loser, CSP capex consensus at $950B FY26, +94% y/y). The counterargument lives inside the bull case itself — GS Privorotsky flags the financing structure ($500bn+ third-party financing platform, ~$366bn future commitments, residual-value support up to 25%) as 'the bigger issue.' Portfolio consequence: own the broadening, not just the leaders; the ROIC evidence gap is the identifiable fragility.

Tension

Record earnings momentum vs. an ROIC evidence base that is almost entirely narrative.

What changed

NVDA delivered above-consensus results and CY27 guidance, and management laid out financing/guarantee structures with enough clarity that the stock reversed an initial dip.

Desk views

GS equity research: buy — CY27 guidance consistent with the most optimistic buyside expectations, margin guidance a clearing event, and commitment transparency reassuring on financial risk.

Evidence IDs · P029

Pasquariello: the interplay between remarkable S&P earnings growth and the enormous AI capex bet leaves the ROIC burden of proof 'huge' — 55% AI-infrastructure growth is very difficult to sustain.

Evidence IDs · P018

Coppersmith: positioning was already de-risked into the print — largest selling since Liberation Day week, net leverage at a 1-year low — making NVDA the test of the growth narrative; AI has become the cyclical trade while ex-AI behaves defensively.

Evidence IDs · P010

Privorotsky: Chinese AI demand is real and showing up in production workloads (MiniMax enterprise revenue +703%, July tokens 20x January), and falling inference costs are unlocking usage — but the ecosystem financing structure is the bigger issue.

Evidence IDs · P014

JPM thematic: tech broadening with dispersion — software no longer trades as a monolithic AI loser, and CSP capex consensus for FY26 is $950B, +94% y/y, with early positive ROI evidence in cloud/AI operating profits.

Evidence IDs · P022

Investor implication

Post-NVDA relief supports the tape near-term, but the durable expression is broadening (software, memory, power infrastructure) rather than concentrated leadership; watch whether the ex-AI market confirms.

Trade expression

Favor dispersion-friendly expressions — software and semis both, rather than the +Semis/-Software pair — while respecting that AI is now the market's primary cyclical expression.

Invalidation

Hyperscaler capex guidance cuts, or a quarter where AI-infrastructure EPS growth decelerates sharply while ROIC evidence remains at the 2% quantification rate.

Verdict

August HICP is expected to be the highest print in three years, and CACIB expects second-round effects from the energy shock to bite essentially during H127 — the ECB's easing story is not clean.

Analysis

CACIB's inflation strategists forecast HICPxT reaching 3.60% y/y in Dec-26 before easing to 3.10% in Jun-27 and 2.06% in Dec-27, roughly in line with market fixings until end-2027. The mechanism is energy passthrough: higher energy prices should progressively bite services inflation, with services rising to 3.4% on average in H127 — most second-round effects materializing next year. The counterweight is that market fixings broadly agree through 2027, so much of this is priced; the divergence risk is in H128. Portfolio consequence: European rate-sensitives face a stickier inflation backdrop than a naive 'ECB cutting cycle' narrative implies, and the energy shock that drives it is the same Strait-of-Hormuz geopolitics showing up across the packet.

Desk views

CACIB: second-round effects from the energy shock kick in progressively after summer and bite in H127; HICPxT forecast raised to 3.60% y/y for Dec-26.

Evidence IDs · P005

Investor implication

Do not position for an early, aggressive ECB easing path; European duration and rate-sensitive equities carry an energy-inflation headwind into 2027.

Invalidation

A rapid, durable Strait of Hormuz normalization that collapses energy prices before H127 would defuse the second-round mechanism.

Verdict

The late-July joint BoJ–US Treasury intervention bought time but is fading; with the BoJ signaling a possible September hike and US drivers still dominant, USD/JPY is likely to remain range-bound near 159-160 rather than trend sharply either way.

Analysis

UniCredit's framing is the sharpest: the joint intervention was 'the strongest response Japan could send to markets to prevent a further plunge of the JPY,' driving a ten-big-figure squeeze from near 164 to 155.24 — yet USD/JPY is back above 159 despite a weaker dollar and renewed BoJ hike expectations. Their verdict: if Japanese authorities only passively limit weakness rather than aggressively drive the yen higher, intervention is 'a losing game,' with the Takaichi government's fiscal plan the potential key driver. UBS agrees on the destination: US factors dominate — a resilient economy and still-hawkish Fed keep the pair supported — and they maintain 160 for Dec-26 and 158 through Sep-27, viewing a substantial sustained yen rebound as low-probability and against policymakers' preferences. MUFG adds the domestic catalyst: Deputy Governor Himino said the BoJ needs to 'pay more attention to upside inflation risks than before' and need not wait for full data on past hikes, with hike pricing above 80% — the closest thing to guidance that the pace of tightening could increase. The tension is real: a hawkish BoJ and fading intervention point one way, US resilience the other. Portfolio consequence: range-bound USD/JPY favors yield-carry structures over directional yen bets, and a September BoJ hike is now a live event for global carry and vol.

Tension

Hawkish BoJ signaling plus intervention history vs. dominant US drivers keeping the pair supported.

What changed

Intervention effects have faded with USD/JPY recovering from 155.24 to above 159, while BoJ hike pricing rose above 80% for September.

Desk views

UniCredit: passive intervention is 'a losing game' — the joint action was less successful than the BoJ's solo 2024 effort; the Takaichi fiscal plan may be the key JPY driver.

Evidence IDs · P034

UBS: US drivers dominate; maintain 160 (Dec-26) and 158 through Sep-27; a disorderly yen appreciation is not something policymakers welcome — favor selling upside above 163 and downside below 155.

Evidence IDs · P031

MUFG: Himino's hawkish tone (upside inflation risks, no need for full data on past hikes) is the closest to guidance for a faster hike pace; NVDA-driven low FX vol supports risk appetite.

Evidence IDs · P026

GS pulled forward its next BoJ hike call to September 2026 as USD/JPY reverted toward its pre-intervention range and the BoJ flagged accommodative financial conditions.

Evidence IDs · P017

Investor implication

Trade the range, not the breakout: yield-enhancement and vol-selling structures fit the consensus view, with a September BoJ hike as the principal event risk to carry.

Trade expression

Range strategies around 155-163 per UBS; avoid directional yen trend bets until the Takaichi fiscal plan or a Fed turn breaks the equilibrium.

Invalidation

A Fed pivot toward cuts (UBS's stated high bar) or aggressive Japanese fiscal/monetary coordination to drive the yen higher would break the range.

Verdict

The BOK hiked back-to-back to 3.00% but guided dovishly (one more hike to 3.25%), while Australia's hot July CPI pushed GS to call a November RBA hike — both point to a periphery that is tightening into global stickiness.

Analysis

Korea: the BOK's 27 August 25bp hike to 3.00% was driven by elevated inflation and semiconductor-cycle income growth, but the dot plot (10 of 21 dots at 3.25%, max 3.50%) signals just one more hike until Q127 — more dovish than markets feared. CACIB sees flattening value in a steep KTB curve and cut USD/KRW forecasts to 1,385 (end-2026) and 1,360 (end-2027). Australia: July headline CPI rose 1.0% m/m with trimmed mean at 3.6% y/y, above both GS and market expectations, with pressure broadening into market services and consumer durables; GS upgraded its 3Q26 trimmed-mean forecast to 0.93% q/q — a material upside surprise to RBA forecasts — and now expects a 25bp November hike to 4.60% with a material September risk. Westpac confirms the market repricing: September hike odds moved from ~10% to ~30%, fully priced by February 2027. The mechanism in both cases is the same: domestic services inflation and, in Korea's case, the semiconductor cycle (which ties back to the AI theme) forcing tightening while the US debate rages. Portfolio consequence: long KRW and AUD carry exposure is supported by hawkish central banks, but Korea's equity market is doubly exposed to the AI cycle's health.

Desk views

CACIB: BOK hiked to 3.00% on semiconductor-cycle strength but guided for only one more hike to 3.25%; KTB curve steepness offers flattening value; USD/KRW forecasts cut to 1,385/1,360.

Evidence IDs · P006

GS Australia: July CPI beat on broadening services and durables pressure; 3Q26 trimmed-mean forecast raised to 0.93% q/q; RBA November hike to 4.60% now most probable, with a material September risk.

Evidence IDs · P015

Westpac: swaps now see a September RBA hike at ~30% (from ~10% earlier in the week), fully priced by February 2027; local yields bear-steepened on the surprise.

Evidence IDs · P035

Investor implication

Hawkish periphery central banks support regional FX carry, but Korea's tightening coincides with its equity market's leveraged exposure to the AI capex cycle — a two-sided risk.

Invalidation

A sharp AI-cycle disappointment would undermine the semiconductor-income channel underpinning the BOK's hawkishness; a Hormuz-driven energy spike would reinforce the RBA's problem.