Weekly Market Outlook: Aug 31 - Sep 4, 2026

A roundup of the research and analysis we went through this week (largely pre-dating Friday's Jackson Hole reaction), organized by theme rather than by name or level where the macro sections below already cover the specifics.

Nvidia's print dominated the pre-read. Revenue came in near $96.2B (+106% YoY) with data-center revenue around $89B, EPS roughly $2.46 (+128% YoY), and guidance for the next quarter of about $108B against consensus closer to 45% growth — guidance the panelists across multiple sources read as evidence Nvidia remains supply-constrained, not demand-constrained. Gross margin held near 75-78%, operating margin 66.2%, and management flagged plans to deploy an additional 2 million GPUs through fiscal 2029. The one yellow flag repeatedly cited: inventory rose to about $31.5B (+22.4% QoQ) and purchase commitments jumped from $119B to $279B, largely memory-related — a bet that demand holds through 2028-2029 rather than a red flag today. Salesforce's post-earnings pop (Anthropic partnership, EPS beat) was read across sources as pushing back on the 'AI kills SaaS' narrative — durable systems of record with embedded enterprise data look more likely to be enhanced by agentic AI than replaced by it, a distinction worth carrying into how we think about software versus infrastructure exposure. Marvell was the counter-example: a genuinely strong quarter (record $2.74B revenue, +37% YoY, FY27 outlook raised to ~$12B from $11.5B) that still fell nearly 10% on valuation/expectations — a reminder that 'good enough' isn't always good enough once the bar embeds a beat.

On the macro/fiscal side, multiple sources converged on the same tension: Treasury's active yield-management (doubled long-bond buybacks, yen intervention, FIMA-facility discussion) is increasingly blurring into territory traditionally seen as the Fed's, and Druckenmiller's public criticism of Bessent's approach was widely discussed as valid on its core point even if the reaction to it was overstated — the deeper problem is a Congress-driven structural deficit that no single Treasury tactic fixes. Hyperscaler debt issuance is now estimated near 9% of investment-grade supply (roughly double a year ago), a genuine crowding-out dynamic for Treasuries that several sources tie directly to this year's stubbornly elevated long-end yields — worth watching now that Warsh has added a hawkish Fed-policy layer on top of that fiscal pressure. The AI-bubble debate itself remains genuinely two-sided: the bull case (two compounding adoption curves — more users, more tokens per user) is real, but the more skeptical read (roughly 100x pricing gaps between frontier and open-weight models, e.g. ~$50/M output tokens versus ~$0.41/M for Qwen) argues token deflation could eventually pressure infrastructure economics even as usage keeps growing — a 6-12-month constructive, 2-5-year cautious framing that several sources landed on independently.

Crypto and BMNR-specific notes: Bitcoin's reclaim of its 200-day moving average on a second attempt (a stronger technical signal than the earlier failed breakout) drove a multi-day rally with roughly $3.1B in short liquidations before Friday's Warsh-driven pullback to $77,678 gave back a chunk of the move — institutional ETF demand stayed intact through it (~$2.8B in net inflows over the eight sessions preceding the speech). BitMine (BMNR) is now holding roughly 5.85M ETH (~4.84% of supply, ~97% of its stated 5% accumulation target) with an mNAV sitting just below 1.0x — meaning the stock currently trades roughly in line with, not at a premium to, its underlying ETH value, and its ~5-month cash runway against obligations still depends on further ETH appreciation, accretive issuance, or staking-reward sales rather than staking income alone. On the physical-AI side, Tesla's Nevada robotaxi cap rose from 10 to 5,000 vehicles (a regulatory unlock, not a deployment number — Tesla itself guided to ~2,500 within 12 months), while Nebius cleared its Vineland, NJ data-center approval after an initial scare and upsized its convertible to $5B at a premium strike — both read as funding/capacity moves rather than distress signals, with institutional ownership reportedly climbing from ~41% to ~78% in under a year.

1. 01SPX closed the prior week at 7,711.71 (-0.25% Friday) after Warsh's hawkish tone hit rate-sensitive corners hardest — the Dow held nearly flat (53,559.99, -0.02%) while the Nasdaq (26,402.42, -0.52%) and especially the Russell 2000 (2,974.72, -1.3%) underperformed, the classic signature of a front-end-driven repricing rather than a broad growth scare.
2. 02The 2-year yield jumped 12bp to 4.35% and the 10-year rose to 4.73% (+5.6bp) — the front end moving more than the long end confirms markets are repricing Fed policy specifically, not a generic term-premium or fiscal-stress move.
3. 03DXY firmed to 99.65 (+0.5%) — a clean reversal of the multi-week dollar-weakness trend that had been building on rate-cut optionality through most of August.
4. 04Gold sold off nearly 3% to roughly $4,530/oz and silver fell over 3% — a sharp unwind of the inflation/geopolitical hedging premium that had been building since early August, consistent with markets pricing a Fed that intends to act on inflation rather than tolerate it.
5. 05Bitcoin fell over 3% to $77,678 even as spot ETF inflows stayed positive through the prior eight sessions (~$2.8B) — a sign the crypto complex is trading Fed-policy-sensitive for now, not purely a structural-adoption story, and a genuine tell for BMNR/MSTR into the week.
6. 06The most vulnerable consensus: CME FedWatch's 57% hike probability is up sharply but still leaves a real 43% hold case — and Polymarket/Kalshi pricing (a tighter 52/48 split) suggests the market hasn't fully converged on Warsh's framing. A soft ADP/ISM/payrolls sequence this week could snap odds back toward hold just as quickly as they moved toward hike.

Secondary threads: Broadcom (AVGO) earnings Wednesday after the close carry a widely cited ~$60B AI backlog and will be read as a direct extension of Nvidia's supply-constrained-not-demand-constrained narrative from the prior week; Zscaler (ZS) reports Thursday as a smaller cloud-security read-through; and the crypto complex (BMNR, MSTR) remains the most Fed-sensitive corner of the market heading into a week where the Fed narrative itself is the dominant catalyst.

The bull case rests on this week's labor data (ADP, JOLTS, claims, and ultimately Friday's payrolls) coming in soft enough to partially unwind Friday's hawkish repricing — echoing July's -23K NFP shock and reminding the Fed that the labor side of its dual mandate is still visibly weakening even as inflation runs hot. In that scenario, hike odds drift back toward 40-45%, yields and DXY give back Friday's gains, gold and Bitcoin stabilize, and equities resume their record-high grind with Broadcom's AI-backlog commentary reinforcing the capex-durability thesis Nvidia set up the prior week. The bear case is that this week's data confirms Warsh's framing: a resilient ADP/ISM Services print and a payrolls number that shows genuine strength (not just a bounce off July's outlier miss) would push hike odds toward 65-70%+ heading into September 16 — a scenario equities have not fully priced given the SPX gave back only 0.25% on Friday's initial 17-point odds swing.

Friday's payrolls report is the single data point most likely to resolve this tension, with Thursday's ADP/ISM Services/claims cluster setting the tone into it. The cross-asset behavior worth trusting: a genuine hold-case rebuild should show the 2-year yield falling faster than the 10-year (front-end-specific, Fed-driven), DXY breaking back below 99, gold and Bitcoin stabilizing or bouncing, and small caps outperforming as their higher rate-sensitivity works in reverse. A genuine hike-case confirmation should show 2-year yields extending their Friday jump, DXY holding or extending above 100, VIX waking meaningfully from its still-low base, and credit spreads widening alongside — not just equities drifting sideways on light volume, which would suggest complacency rather than a resolved read.

The fiscal backdrop from the 'What We've Read' section — hyperscaler debt issuance running near 9% of investment-grade supply, roughly double a year ago — is worth holding in view here too: even a dovish resolution to this week's labor data doesn't remove the structural upward pressure on long-end yields from AI-capex-linked corporate borrowing, meaning the 10-year and 30-year may not fall as cleanly as the 2-year even in the bull scenario.

1. 01CME FedWatch prices the September 16 FOMC decision at roughly 57% hike / 43% hold as of the weekend, up from ~40% hike odds on August 21 — a 17-point swing driven entirely by Warsh's Friday keynote. Polymarket briefly spiked to 68% hike before settling; Kalshi shows a tighter ~48% hike / 52% hold split, underscoring genuine disagreement about how much weight to put on one speech.
2. 02Warsh's core argument: core PCE inflation ran 3.7% over the trailing 12 months and 4.1% over the trailing 6 months, both well above the Fed's 2% target, which he called a 'firm, fixed target' — a materially more hawkish framing than the 'big picture, non-committal' tone he had signaled in his own July 29 comments ahead of the speech.
3. 03This directly extends the July 29 FOMC's three-way regional-president dissent in favor of a hike (the first three-way same-direction dissent since September 2016) — Warsh's speech reads as siding with those dissenters rather than the July hold decision itself.
4. 04With the front end (2-year, +12bp to 4.35%) moving more than the long end (10-year, +5.6bp to 4.73%) on Friday, this week's data is best read through a Fed-specific lens first — a further front-end move on hot labor data would confirm the hike case is strengthening; a front-end reversal on soft data would confirm markets are still genuinely undecided.
5. 05Key signals to watch through the week: 2-year yield direction (the cleanest Fed-specific read), whether DXY holds above the 99 support level it broke through Friday, whether VIX begins pricing real event risk ahead of Friday's payrolls rather than only reacting after the fact, and whether Thursday's ADP/ISM Services cluster moves hike odds further before payrolls even print.

Equity moves are more credible this week when Treasury yields and the dollar confirm the same read — a stock rally on soft labor data should show 2-year yields and DXY falling together with gold/Bitcoin stabilizing, not just an equities bounce on light volume that ignores what the rates market is doing.

1. 01Mega-cap tech/AI: Broadcom's Wednesday print is the group's main swing factor — a ~$60B AI backlog holding up would extend the Nvidia-driven capex-durability thesis from the prior week; a disappointment lands awkwardly on top of an already-hawkish macro tone.
2. 02Semis/AI infrastructure: dispersion remains the more interesting story than direction post-Nvidia — names with defensible data, distribution, or scarce physical/power inputs (echoing the multiple-compression thesis from this week's research) look better positioned than broad-basket AI exposure if capex-fatigue questions resurface.
3. 03Small caps/Russell 2000: the single most Fed-sensitive group in the index, and it showed — down 1.3% Friday versus SPX's-0.25%. Small caps are the cleanest tell for how this week's labor data is being read: sharp outperformance on soft data, sharp underperformance if hike odds extend further.
4. 04Financials: a front-end-led yield move (2-year outpacing 10-year) is generally less constructive for banks than a steepening curve driven by growth optimism — watch whether the curve bull-flattens (Fed-driven, growth-scare-adjacent) or steepens (term-premium/growth-driven) as the week's data lands.
5. 05Crypto-adjacent equities (BMNR, MSTR): the most directly Fed-sensitive names in this corner of the market right now — Bitcoin's >3% Friday drop despite eight straight sessions of ETF inflows shows the asset is trading macro-sensitive for now, and BMNR's mNAV sitting just below 1.0x means near-term equity performance leans more on Fed-driven ETH price action than on the underlying accumulation story this week.
6. 06Consumer/defensives: no major consumer print this week, but any rotation into staples/utilities alongside a further yield/DXY move would be the tell that markets are reading this as more than a routine Fed repricing — worth watching as a secondary signal even without a scheduled catalyst.
7. 07Software: Zscaler's Thursday print is a smaller, secondary test of the Salesforce-driven 'SaaS isn't dying to AI' thesis from the prior week's research — a clean beat would add a second data point to that narrative.

ADP, ISM Services, and Friday's payrolls all print close to or below consensus, echoing July's labor-market softness and giving the Fed room to hold on September 16 despite Warsh's hawkish framing; hike odds drift back toward 40-45%, the 2-year and DXY give back most of Friday's move, gold and Bitcoin stabilize, and equities resume grinding toward highs with Broadcom's AI-backlog commentary reinforcing capex durability. Invalidation: any single print in the ADP/ISM/NFP sequence surprising meaningfully to the upside, or Broadcom guidance disappointing badly enough to compound the macro tension.

This week's labor sequence comes in soft enough that hike odds recede meaningfully from 57%, Broadcom's print extends the Nvidia-driven AI-capex thesis without a scare, and Friday's cross-asset move (yields, DXY, gold, Bitcoin) largely retraces. Cross-asset confirmation requires the 2-year yield falling faster than the 10-year, DXY breaking back below 99, small caps outperforming as their rate-sensitivity works in reverse, and gold/Bitcoin stabilizing rather than continuing to fall. False-breakout risk: a rally driven only by a single soft print (e.g., weak claims) without ADP, ISM Services, and payrolls all confirming would be a partial, not durable, unwind of Friday's repricing.

ADP and ISM Services both come in resilient, claims stay low, and Friday's payrolls show genuine strength rather than a bounce off July's outlier miss — pushing hike odds toward 65-70%+ heading into the September 16 FOMC. Expected behavior: the 2-year yield extends its Friday jump, DXY holds or breaks above 100, small caps underperform sharply again, VIX wakes meaningfully from its still-low base, and gold/Bitcoin extend Friday's declines as the inflation/debasement hedge narrative loses urgency to a genuinely tightening Fed. Stabilization would likely require either a dovish clarifying comment from another Fed official in the days after, or Broadcom's earnings providing enough of an offsetting positive catalyst to keep equities from a deeper, broader selloff.

This week concentrates the highest-stakes labor-data sequence of the quarter (ADP, ISM Services, claims, and payrolls) directly on top of a Fed narrative that just swung 17 points in a single speech, plus a major single-stock catalyst (Broadcom) landing mid-week — against a VIX that, even after ticking up off cycle lows Friday, has not remotely priced a scenario where hike odds push toward 65-70%. For a typical SPX/index credit-spread and iron-condor approach, this argues for reduced size, wider strikes, and defined risk positioned to survive the full Tuesday-through-Friday data sequence rather than carrying full-size short premium through any single leg of it — Friday's Warsh reaction already demonstrated that this market can move meaningfully on a single data point when the Fed's reaction function is this actively contested, and the reward for selling cheap volatility into a payrolls print with genuine two-way outcomes does not obviously compensate for that tail risk.