The Productivity Surprise Arrives — Unit Labour Costs Halve the Forecast, but Oil Rebounds and the Dow Snaps Its Record Streak
Data:
Main Theme: “The Best Data of the Year, and Stocks Still Fell” — Q2 productivity came in at 1.4% against 0.6% expected and unit labour costs at 1.3% against 2.1% — the single most benign inflation combination of 2026. Equities fell anyway, as crude rebounded, Treasury yields rose and the Dow snapped a five-session record streak with a 464-point decline.
Thursday delivered exactly the outcome flagged as the bull case in the previous edition, and the market did not take it. Nonfarm productivity rose 1.4% in Q2 against a 0.6% consensus — more than double — while unit labour costs rose just 1.3% versus the 2.1% expected. With compensation running at 3.4%, that is the arithmetic that says wage growth is being absorbed by output rather than passed into prices. Challenger layoffs fell to 33,429 in July, the fewest in two years, and jobless claims came in below consensus at 199,000.
And yet the Dow fell 464.02 points (−0.85%) to 53,885, breaking a five-day winning run of record closes. The reason was mechanical rather than macro: crude rebounded, energy was the only sector to rise, and Treasury yields moved higher. Positioning into Friday’s payrolls did the rest.
🟥 U.S. Equities | The Record Streak Ends
| Index | Closing Level | Net Points Change | % | Session Stance |
| Dow Jones Industrials | 53,885 | 🟥 −464.02 | −0.85% | Snapped a five-day record streak; Salesforce −3% a key drag |
| S&P 500 | 7,709.96 | 🟥 −13.59 | −0.18% | Second consecutive decline |
| Nasdaq Composite | 26,348.35 | 🟥 −15.55 | −0.06% | Best relative performer; QQQ briefly broke its 50-day before recovering |
| Sector breadth | — | — | — | Energy was the only sector higher |
The decliners were concentrated in disappointing guidance rather than weak results: Salesforce −3% on a leadership shuffle, weighing directly on the Dow; SanDisk down 6–9% as Wednesday night’s guidance continued to be sold; AppLovin −14% to −20% on a top-line miss and soft Q3 outlook; Datadog fell roughly 17% to the mid-$230s despite beating on revenue and earnings and raising full-year guidance — after nearly doubling in 2026; Fiserv −12% after cutting its 2026 adjusted EPS outlook to $7.20–7.40 from $8.00–8.30; Peloton −13% on weak fiscal 2027 guidance; Cedar Fair −18% on a revenue and EBITDA miss; and Honeywell spinoff Solstice Advanced Materials −19% in its first report as an independent company, cutting organic sales growth guidance to 4–5% from 7–9%.
The earnings scorecard remains exceptional. Of the 411 S&P 500 companies reported so far, 87% have beaten estimates against 82% a year ago, with the average beat at 14% and average year-over-year earnings growth of 25% versus 14% last year. The problem is not results; it is that beats are now fully priced.
Two macro-relevant warnings landed during the session. JPMorgan CEO Jamie Dimon cautioned that leverage across financial markets remains unusually high, arguing investors may be underestimating risks tied to borrowing outside traditional margin accounts. St. Louis Fed President Musalem said inflation expectations remain anchored near 2% but warned conditions are fertile for them to become unanchored, describing financial conditions as very accommodative with elevated asset prices.
🟥 Asian Bourses | Memory Gives Back Wednesday’s Rally
Asia reversed sharply, trading Wednesday night’s US pullback and the SanDisk and Western Digital guidance disappointments.
South Korea (KOSPI): approximately −2%, with SK Hynix down more than 5%, surrendering a large part of Wednesday’s 5.77–6.72% gain.
Japan (Nikkei 225): −617 points to 65,683, roughly −0.9%. Kioxia slumped 10% — the sharpest regional decline — a direct read-through from SanDisk’s guidance, given the two operate a NAND manufacturing joint venture in Japan.
China / Hong Kong: Shanghai Composite +21 to 3,900; Hang Seng −385 to 25,530. Asian markets closed lower across the board with the exception of India’s Nifty 50 and Australia’s ASX 200.
Europe held up better: the German DAX rose 62 points to 26,189 and the FTSE 100 added 32 points to 10,920.
The pattern is now clear and worth stating plainly. Korean and Japanese memory names have moved 5–10% in alternating directions on four of the last five sessions, driven entirely by US after-hours guidance from SanDisk, Western Digital and AMD. This is no longer a Korean leverage story — the leveraged ETF volumes have collapsed under regulatory pressure — it is a pure guidance-transmission story.
🟧 Commodities | Crude Rebounds on Deal Scepticism
Oil rose and was the session’s only sector winner in equities. WTI gained around 1.3% to $76.20, with Brent trading near $79.91–80.00.
The reversal came from the detail of the proposed deal, not from a breakdown. Reuters reported that an Oman-brokered agreement could hand Tehran control over inbound shipping through the Strait of Hormuz — a potentially major concession in the five-month conflict. Iran said the shipping agreement was in its final stages, but:
- Significant sticking points remain over control of shipping and cargo fees, with Washington continuing to resist Iranian authority over the waterway.
- Some Iranian sources rejected Trump’s claim that an agreement was imminent, saying important details remained unresolved.
- Tehran is demanding the lifting of the US blockade of Iranian ports as part of any arrangement.
The market read this as a fifth false dawn since February. After roughly $11 came out of Brent across Monday and Tuesday, traders were unwilling to price further downside on an agreement whose central term — who controls the waterway — is exactly what remains contested.
Gold traded around $4,267/oz, up roughly 0.5%, extending its move to multi-week highs. The dollar index was essentially flat at 99.75, up 0.08%.
📰 Macro “Red News” | The Bullish Print Nobody Was Positioned For
This was the most important data release of the week, and it broke decisively in the disinflationary direction.
| Indicator | Actual | Consensus | Read |
| Q2 Nonfarm Productivity | +1.4% | +0.6% | More than double expectations; Q1 revised up to +0.8% from +0.3% |
| Q2 Unit Labour Costs | +1.3% | +2.1% | Well below forecast; Q1 revised down to +1.3% from +1.8% |
| Initial Jobless Claims (wk Aug 1) | 199,000 | 202,000–205,000 | Up 1,000 from 198,000; 4-week average fell to 198,750 from 203,250 |
| Continuing Claims (wk Jul 25) | 1.801m | 1.790m | Up from 1.777m — modest softening |
| Challenger Job Cuts (July) | 33,429 | — | Fewest in two years; −27% MoM, −46% YoY |
Why this matters more than any other print this week. The Fed’s inflation problem has been a cost problem. The Q2 Employment Cost Index showed compensation running at 3.4% year-over-year, and ISM Services prices hit 70.3 on Wednesday with a twelve-month average at a three-year high. The mechanism that turns 3.4% compensation growth into consumer inflation is unit labour costs = compensation growth minus productivity growth.
At 1.4% productivity, unit labour costs came in at 1.3% — roughly consistent with a 2% inflation target. This is the first substantive evidence that the AI capital expenditure cycle is showing up in measured output per hour. It also directly undercuts the argument made by the three July FOMC dissenters — Hammack, Kashkari and Logan — that a 25bp hike is necessary.
The market repriced accordingly, if modestly. CME FedWatch showed the probability of a September hike at roughly 54.9%, down from about 65% on Tuesday. The 10-year traded around 4.62% and the 2-year at 4.20% early, though yields rose through the session as crude rebounded.
The one caveat worth holding. Preliminary productivity data is heavily revised, and a single quarter does not establish a trend — Q1 was revised up from 0.3% to 0.8%, which is a large adjustment in its own right. Confirmation matters more than the initial print.
🌙 After the Bell | Consumer and Cloud Deliver
Airbnb (ABNB) closed at $151.64 and surged 7–9% after hours to around $164.48.
- Q2 EPS of $1.37 on revenue of $3.61 billion, against $1.25 and $3.58 billion expected. Revenue rose 17%; adjusted EBITDA rose 21% with the margin expanding a point to 35%.
- Full-year revenue and margin guidance were raised, with annual revenue now projected to grow at least in the mid-teens percentage and the minimum margin lifted to 35.5%.
- The detail worth flagging: customer-support spending per booking fell roughly 16%, which Airbnb attributed to its AI assistant. Sales and marketing rose 27% to $875 million against 17% revenue growth — the company is deliberately spending the AI efficiency gain on expansion. This is one of the cleanest examples yet of AI showing up in an operating cost line rather than a capex line, and it arrived on the same day as the productivity print.
Cloudflare (NET) jumped 16–17% on strong guidance: Q3 adjusted EPS of $0.34 on revenue of $736–737 million, against consensus of $0.32 and $722 million, with Q2 beating on both lines.
Others: Instacart (Maplebear) +8% on a revenue beat despite an EPS miss; Lyft marginally higher with revenue of $1.84 billion beating $1.81 billion but EPS of $0.13 missing $0.14; DraftKings −1.5% on a revenue miss of $1.44 billion against $1.51 billion and a $0.14 loss per share where the Street expected a $0.02 profit, though full-year guidance was reaffirmed.
📌 Reading the Session
- The data was the best of the year and the tape ignored it. Productivity at 1.4% with unit labour costs at 1.3% is the combination that resolves the Fed’s dilemma. Equities fell because oil rose and yields rose — a positioning outcome, not an information outcome. That gap is worth watching.
- The Hormuz trade has hit its scepticism threshold. The proposed Oman-brokered terms would give Tehran control over inbound shipping — the exact point Washington resists. After $11 of Brent has already been priced out, the market now requires signature, not headlines.
- “Beat and raise” has now failed four consecutive nights. AMD, then SanDisk and Western Digital, then Datadog — which beat, raised, and fell 17% after nearly doubling year-to-date. With 87% of reporters beating, the beat is not the variable; the prior move is.
Friday: July nonfarm payrolls at 19:30 ICT. Consensus is roughly 80,000–85,000 with unemployment expected to hold at 4.2% and average hourly earnings at 3.5% year-over-year. June printed +57,000. The forecast range is unusually wide at 40,000 to 157,000.
Companies
Theme: “Beat, Raise, Fall — Four Nights Running” — Datadog beat on revenue and earnings, raised full-year guidance, and fell 17%. It had nearly doubled in 2026. Meanwhile Airbnb and Cloudflare — neither of them a market darling this year — surged on results that were good rather than spectacular. The market is not rewarding performance; it is repricing expectations.
Thursday completed a four-session sequence with no exceptions. AMD, SanDisk, Western Digital and now Datadog have each beaten consensus and been sold hard. Each had risen more than 100% year-to-date. Against that, Airbnb, Cloudflare, Palantir and Disney — all trading well below their highs or below their year-start levels — have each surged on results that were solid rather than extraordinary. With 87% of 411 S&P 500 reporters beating estimates and average earnings growth running at 25%, the beat conveys no information. The prior move does.
📉 1. Datadog: The Purest Example Yet
Datadog (DDOG) fell roughly 17% to the mid-$230s after beating on Q2 revenue and earnings and raising full-year guidance.
There is no fundamental complaint available here — the company delivered on every axis investors claim to care about. The stock had nearly doubled in 2026 heading into the print. That is the entire explanation, and it is the same one that applied to AMD (+140% year-to-date, fell 8% on 107% data-centre growth), SanDisk (+400%, fell 10% on a 12% EPS beat) and Western Digital (+220%, fell 10% on a beat-and-raise).
The observability and monitoring layer matters strategically — Datadog is a direct beneficiary of AI workload proliferation, since more inference means more to monitor. The de-rating is about multiple compression in a crowded position, not about the thesis.
🏠 2. Airbnb: AI in the Cost Line, Not the Capex Line
Airbnb (ABNB) closed at $151.64 and rose 7–9% after hours toward $164.48, describing its “strongest results in years.”
| Metric | Q2 2026 | Versus expectation / prior |
| EPS | $1.37 | vs $1.25 expected |
| Revenue | $3.61bn, +17% | vs $3.58bn expected |
| Adjusted EBITDA | +21% | Margin expanded 1pt to 35% |
| Sales and marketing | $875m, +27% | Deliberately outrunning revenue growth |
| Customer support cost per booking | −16% | Attributed to the AI assistant |
| FY guidance | Revenue growth at least mid-teens % | Raised; minimum margin lifted to 35.5% |
Why this quarter is more interesting than its headline. Almost every AI datapoint this earnings season has been a spending datapoint — hyperscaler capex now guided at $720–745 billion collectively, SpaceX at $18.37 billion a quarter, AMD tripling capex to $808 million. Airbnb reported an AI datapoint on the cost side: support spend per booking down 16%. That is measurable operating leverage from deployment rather than from construction.
The timing is notable. It landed on the same day Q2 productivity printed at 1.4% against a 0.6% forecast. One company does not make a macro trend, but the mechanism the productivity data implies — output per hour rising because of technology deployment — is exactly what Airbnb described in its cost structure. Management then chose to reinvest the gain: sales and marketing up 27% against revenue up 17%.
☁️ 3. Cloudflare: Guidance Does the Work
Cloudflare (NET) jumped 16–17% after guiding Q3 to adjusted EPS of $0.34 on revenue of $736–737 million, against an LSEG consensus of $0.32 and $722 million. Q2 also beat on both lines.
The read-through matters for the AI application layer: Cloudflare sits at the network edge for a large share of inference traffic, and forward guidance above consensus suggests that traffic is still compounding. Together with Palantir’s 93% revenue growth on Monday, the software and edge layer is producing the cleanest fundamental acceleration in the complex — and, crucially, doing so without the capital intensity that has punished the hardware names.
💾 4. The Memory Complex Extends Its Decline
SanDisk fell a further 6–9% during Thursday’s session, adding to Wednesday’s 5.4% regular-session loss and after-hours decline. The market continues to focus on the composition of its quarter rather than its size: revenue of $8.97 billion beat by 5.7% and adjusted EPS of $39.25 beat by 12.3%, with an 84.6% gross margin — but roughly $2.01 billion of the sequential revenue gain came from pricing, and Q1 guidance of $10.3–10.8 billion underwhelmed at the midpoint.
Western Digital also traded lower despite results that topped estimates and guidance above consensus, with investors focused on decelerating sequential improvement.
The transmission into Asia was immediate and severe. Kioxia slumped 10% in Tokyo — it manufactures NAND jointly with SanDisk in Japan — and SK Hynix fell more than 5% in Seoul, giving back most of Wednesday’s gain. This is now a clean guidance-transmission mechanism rather than the leverage story of late July; Korean single-stock leveraged ETF volumes have already collapsed under regulatory pressure.
The date to mark: SanDisk’s Investor Day on 13 August, covering fiscal 2027 supply, contract conversion and price floors. That is the event that determines whether the 84.6% margin is a cycle peak or a new baseline.
🛰️ 5. SpaceX: The Lockup Arrives
SpaceX (SPCX) rose 2.84% to around $111.34 on Thursday — the day its first lockup expired, making up to 911.5 million insider shares eligible for sale against a public float below 280.1 million.
Context that needs correcting from my previous edition: SPCX closed Wednesday at $108.27, down more than 13% — an all-time low — not the “more than 10%” I reported. It listed on 12 June at $135, reached a record $225.64 on 16 June, and is now roughly 49% below that peak.
The lockup is a multi-month process, not a single event. This first tranche releases 20% of restricted holdings, with smaller tranches continuing through October, a second large tranche after Q3 earnings, and the full backstop expiring 8 December 2026. A structural detail worth knowing: the schedule allowed an additional 10% to unlock early if the stock had traded 30% above the IPO price on five of ten consecutive days before the first earnings release — that condition was not met.
Venture investor Paul Kedrosky has noted that many holders have pledged stock against purchases and need to sell; short sellers have been positioning ahead of the date. University of Florida economist Jay Ritter — who has studied lockups extensively — expects downward pressure but notes not all unlocked shares reach the market. Historical precedent is genuinely mixed: Facebook’s largest expiration in November 2012 closed 12.5% higher; Uber’s 2019 expiry coincided with a 17% decline into the date.
Separately, SpaceX purchased $295 million of Tesla Megapacks for its Texas chip facility — a related-party transaction worth noting for governance-sensitive clients.
📋 6. The Guidance Casualties
- Fiserv (FI) −12% after cutting its 2026 adjusted EPS outlook to $7.20–7.40 from $8.00–8.30.
- AppLovin (APP) −14% to −20% on a Q2 top-line miss and Q3 adjusted EBITDA guidance of $1.71–1.74 billion against a $1.75 billion consensus.
- Peloton (PTON) −13% — it beat on fiscal Q4 earnings and reported its first full year of net profit and operating income, but guided fiscal 2027 revenue to $2.3–2.4 billion against $2.42 billion expected, with connected fitness subscriptions falling roughly 267,000 (−9.8%) as it laps price increases.
- Cedar Fair (FUN) −18% on Q2 revenue of $864.9 million against $933.3 million expected, adjusted EBITDA of $243.1 million against $282.3 million, and a $202.6 million net loss.
- Solstice Advanced Materials (HONA) −19% in its first report since spinning out of Honeywell, cutting 2026 organic sales growth guidance to 4–5% from 7–9% and full-year adjusted EPS to $7.60–7.90 against $8.86 expected, after Q2 adjusted profit fell 32%.
- Salesforce (CRM) −3% on a leadership shuffle — a meaningful drag on the price-weighted Dow.
- DraftKings −1.5% after hours on a revenue miss and an unexpected loss, though full-year adjusted EBITDA and revenue guidance were reaffirmed.
📌 Analyst Take
The tradeable pattern of this earnings season is now unambiguous and worth stating as a rule: with 87% of reporters beating and average earnings growth at 25%, a beat is the base case, not a catalyst. What moves a stock is the distance between the result and the position.
| Company | YTD entering the print | Reaction |
| AMD | +140% | −8% on +107% Data Center growth |
| SanDisk | +400%+ | −10% on a 12% EPS beat, 84.6% margin |
| Western Digital | +220%+ | −10% on a beat-and-raise |
| Datadog | Nearly doubled | −17% on a beat-and-raise |
| Palantir | −29% | +29.45% |
| Disney | −12% | +3.83% |
| Airbnb | Below highs | +7–9% after hours |
| Cloudflare | +42% | +16–17% after hours |
The second, more durable theme is where AI is now showing up. For four months the AI datapoints have all been on the spending side. Thursday produced two on the return side: Airbnb’s support cost per booking down 16%, and Q2 productivity at 1.4% against a 0.6% forecast. If that pattern extends into Q3 reporting, the entire framing of the AI trade changes — from “who can afford the capex” to “who is extracting the operating leverage.” That would favour the application and services layer over the silicon layer, which is precisely the rotation the tape has been running all week.
Friday is a data day rather than an earnings day. The remaining large reports are thinning out; payrolls at 19:30 ICT is the session. Next week brings CoreWeave and Super Micro on 11 August, SanDisk’s Investor Day on 13 August, and July CPI on 12 August.
General
Thursday, August 6th, 2026: The Data Said One Thing and the Tape Said Another
Thursday produced the single most benign macro combination of 2026 and equities fell. Q2 productivity came in at 1.4% against 0.6% expected; unit labour costs at 1.3% against 2.1%; Challenger layoffs at a two-year low; jobless claims below consensus. That is the configuration that dissolves the Fed’s dilemma — it says compensation growth of 3.4% is being absorbed by output rather than passed into prices.
The Dow still fell 464 points and snapped a five-session record streak. The proximate causes were mechanical: crude rebounded on scepticism about the Hormuz terms, Treasury yields rose, and nobody wanted risk into Friday’s payrolls. But the gap between the information and the price action is itself the most interesting feature of the session, and it will resolve one way or the other on Friday.
- Why the Productivity Number Is the Week’s Real Story
The Fed’s inflation problem this year has been a cost problem, not a demand problem. The chain is straightforward:
Compensation growth − productivity growth = unit labour cost growth → core services inflation.
The inputs coming into Thursday were uncomfortable. The Q2 Employment Cost Index showed compensation at 3.4% year-over-year with real wages down 0.4%, the first negative reading since Q4 2022. ISM Services prices hit 70.3 on Wednesday, a twelve-month average at the highest since April 2023. Core PCE is stuck at 3.3%. Three FOMC members — Hammack, Kashkari and Logan — dissented in favour of a 25bp hike at the 29 July meeting, the Fed’s most divided decision since 2016.
Thursday broke the chain at the productivity link. At +1.4% productivity, unit labour costs came in at just 1.3% — a level broadly consistent with 2% inflation. Q1 was also revised up, from +0.3% to +0.8%. For the first time this cycle, there is measured evidence that the AI capital expenditure boom is producing output per hour rather than only depreciation.
The market repriced, but only partially. September hike odds fell to roughly 54.9% from about 65% on Tuesday — a meaningful move, but well short of what a clean unit-labour-cost print would justify in isolation. Two reasons for the hesitation: preliminary productivity data is heavily revised, and one quarter is not a trend.
The corroborating microeconomic evidence arrived the same evening. Airbnb reported customer support cost per booking down roughly 16%, attributed directly to its AI assistant — and then reinvested the saving, growing sales and marketing 27% against revenue up 17%. That is the productivity mechanism visible at the company level.
- The Labour Market Is Not Breaking — It Is Standing Still
A full week of labour data now allows a confident characterisation.
| Indicator | Latest | Signal |
| JOLTS openings (Jun) | 7.359m vs 7.400m exp | Slightly soft; hires, quits, layoffs all unchanged |
| ISM Mfg employment (Jul) | 52.8 | First expansion in ~3 years |
| ISM Services employment (Jul) | 47.4 | Lowest since March |
| ADP private payrolls (Jul) | +44,000 vs +65,000 exp | Weakest since January |
| Challenger job cuts (Jul) | 33,429 | Fewest in two years; −27% MoM, −46% YoY |
| Jobless claims (wk Aug 1) | 199,000 vs 202–205k exp | 4-week average fell to 198,750 |
| NFP consensus (Fri) | ~80,000–85,000 | June was +57,000; range 40k–157k |
The apparent contradiction — weak hiring, record-low layoffs — is the defining feature of this labour market. Firms are neither adding nor cutting. Quits are at 2.0%, meaning workers are not moving either.
The supply side complicates the read considerably. The unemployment rate has stayed low largely because labour force participation has declined; the employment level in 2026 has actually fallen by 833,000. That is why Citigroup holds a well-out-of-consensus call for three rate cuts between now and January 2027, with economist Veronica Clark arguing the unemployment rate rises above 4.5% within a few months. If she is right, the entire September hike debate is misframed.
For Friday, the composition matters more than the headline. A print near consensus with participation rising and unemployment at 4.2% is benign. The same print with participation falling again is not — it would mean the low unemployment rate is an artefact of people leaving the workforce rather than a sign of strength.
- The Hormuz Trade Hits Its Credibility Limit
Crude rose on Thursday for the first time this week — WTI +1.3% to $76.20, Brent near $79.91–80.00 — and energy was the only equity sector to advance.
The reversal came from the substance of the reported terms, not from a collapse in talks. Reuters reported that the Oman-brokered proposal could hand Tehran control over inbound shipping through the Strait — a major concession that Washington continues to resist. Iran said the agreement is in its final stages while some Iranian sources rejected Trump’s claim that a deal was imminent, citing unresolved questions over control and cargo fees, and Tehran is demanding the lifting of the US blockade of Iranian ports.
The asymmetry has now fully inverted from Monday. At the start of the week, a deal was upside and no-deal was neutral. After roughly $11 came out of Brent, a deal is priced and the unresolved terms are the risk. Critically, the contested point is not procedural — who controls the waterway is the entire substance of the five-month conflict.
Practical implication for energy-sensitive exposure: this is now a binary policy trade with a headline-driven distribution, not an earnings trade. Refining margins in particular remain priced on a closed strait.
- Two Warnings Worth Recording
Two senior voices raised the same underlying concern on Thursday, from different directions.
JPMorgan CEO Jamie Dimon cautioned that leverage across financial markets remains unusually high, and that investors may be underestimating risks tied to borrowing outside traditional margin accounts. The specific concerns circulating alongside this are elevated equity valuations, hedge fund leverage, and large Treasury basis trades that could amplify volatility during stress.
St. Louis Fed President Musalem said inflation expectations remain anchored near 2% but warned conditions are fertile for them to become unanchored, describing financial conditions as very accommodative with elevated asset prices and stating that the Fed cannot be driven by markets.
Why these belong together. Both describe the same configuration: asset prices at records, leverage elevated, and policy not yet restrictive enough to lean against it. That is the argument the three FOMC dissenters are making, and Thursday’s good productivity data does not answer it — a benign unit labour cost print addresses the inflation channel but not the financial-conditions channel.
- Asia: A Pure Guidance-Transmission Market Now
The KOSPI fell roughly 2% with SK Hynix down more than 5%, and the Nikkei dropped 617 points to 65,683 with Kioxia slumping 10%.
The mechanism has changed materially from two weeks ago. In late July, Korean volatility was a leverage story — retail single-stock leveraged ETFs, repeated circuit breakers, a 22.2% monthly drawdown and a record 17.91% single-day rebound. Regulatory intervention has now visibly drained that leverage, with SK Hynix-linked leveraged ETF volumes at their lowest since early June.
What remains is direct guidance transmission. Kioxia manufactures NAND jointly with SanDisk in Japan; SanDisk guided softly on Wednesday night; Kioxia fell 10% on Thursday. Wednesday’s Asian rally came from the SK Hynix–SanDisk HBF standards announcement. The same two companies now drive the region in both directions, on a one-session lag.
For portfolio purposes this is an improvement — a guidance-driven market is analysable; a leverage-driven one is not. But it also means Asian memory exposure is effectively a leveraged position on US after-hours guidance, and should be sized accordingly.
📊 Global Macro Sentiment Summary — Thursday, August 6th, 2026
| Narrative Channel | Core Fundamental Trigger | Net Portfolio Posture |
| Index Structure | Dow −0.85% to 53,885, snapping a five-day record streak; S&P −0.18%; Nasdaq −0.06% | 🟨 Consolidation |
| Productivity / costs | Q2 productivity +1.4% vs +0.6% exp; unit labour costs +1.3% vs +2.1% | 🟩 Most benign print of 2026 |
| Labour | Claims 199k, below consensus; Challenger 33,429, fewest in two years; continuing claims up to 1.801m | 🟨 Frozen, not deteriorating |
| Fed pricing | September hike odds ~54.9%, down from ~65% Tuesday | 🟩 Easing |
| Fed commentary | Musalem: expectations anchored but conditions fertile to unanchor; financial conditions very accommodative | 🟥 Hawkish undertone |
| Energy | WTI +1.3% to $76.20; Brent ~$80; energy the only sector higher | 🔄 Deal scepticism |
| Geopolitics | Oman proposal would give Tehran control of inbound shipping; US resists; Iran denies deal imminent | ⚠️ Contested core term |
| Earnings | 411 reported, 87% beat, avg growth 25% — yet DDOG −17% on a beat-and-raise | 🟥 Positioning risk |
| After hours | Airbnb +7–9%, Cloudflare +16–17% — both from non-crowded positions | 🟩 Rotation intact |
| Asia | KOSPI ~−2%, SK Hynix −5%; Nikkei −617 to 65,683, Kioxia −10% | 🟥 Guidance transmission |
| Systemic | Dimon: market leverage unusually high, risks underestimated outside margin accounts | ⚠️ Tail risk flagged |
Upcoming News
Friday, August 7th, 2026 — Theme: “Payrolls Decides the September Argument” — The July employment report is the last major labour input before the 15–16 September FOMC, arriving with hike odds at roughly 55%, a forecast range spanning 40,000 to 157,000, and a labour force that has shrunk by 833,000 this year.
Friday is a single-event session. July nonfarm payrolls land at 19:30 ICT, one week after the Fed’s 9–3 decision to hold at 3.50–3.75% — its most divided vote since 2016, with regional presidents Hammack, Kashkari and Logan all preferring a 25bp hike. Thursday’s productivity and unit labour cost data already weakened the hawkish case materially. Payrolls determines whether it survives at all.
🔴 High-Impact Calendar — Friday, August 7th, 2026
Times in ICT (Hanoi). ET is ICT minus 11 hours.
| Time (ICT) | Currency | Event / Indicator | Consensus | Impact |
| Morning | CNY | China Trade Balance (July) — exports and imports | — | 🟠 Med |
| Morning | JPY | Japan Household Spending | — | 🟢 Low |
| Morning | EUR | Germany Trade Balance and Industrial Production | — | 🟠 Med |
| 19:30 | USD | July Nonfarm Payrolls | ~80,000–85,000 (prior +57,000) | 🔴 High |
| 19:30 | USD | Unemployment Rate | 4.2% (unchanged) | 🔴 High |
| 19:30 | USD | Average Hourly Earnings (YoY) | 3.5% (prior 3.5%) | 🔴 High |
| 19:30 | USD | Average Hourly Earnings (MoM) | ~+0.3% | 🟠 Med |
| 19:30 | USD | Labour Force Participation Rate | ~61.6% (prior 61.5%) | 🔴 High |
| During session | USD | Fed speaker | — | 🟠 Med |
| End of day | — | Weekly BTC / ETH options and futures expiry (Deribit, CME) | — | 🟠 Med |
- The Forecast Distribution Is Unusually Wide
Consensus clusters at 80,000–85,000, but the published range runs from 40,000 to 157,000 — an exceptionally wide dispersion that reflects genuine disagreement about what the recent data means.
| Source | July NFP | Rationale |
| Market consensus | 80–85k | Unemployment 4.2%, AHE 3.5% YoY |
| Wells Fargo (Bill Adams / Fifth Third) | 90–95k | H1 averaged 92k; small business hiring plans improved; claims fell between survey weeks |
| Continuum Economics | 120k | Rebound in leisure and hospitality after a 61k June decline; unemployment to 4.3% |
| Kiplinger panel (one member) | 75k | Unemployment rising to 4.3% |
| Deutsche Bank (earlier in week) | 65k | Modest improvement on June’s 57k |
The June baseline is the key context. June printed +57,000, a sharp slowdown from the preceding three months, explained largely by a 61,000 decline in leisure and hospitality — surprising to those who expected the World Cup to lift the sector, and possibly a seasonal adjustment artefact. Continuum expects a 25,000 rebound in that sector alone. May was revised down to +129,000 from +172,000.
- What Actually Matters in the Release
The headline number is the least informative part of this report. Three other elements carry more weight:
Labour force participation. The unemployment rate has stayed low largely because participation has fallen — the employment level in 2026 has declined by 833,000. Consensus looks for participation to tick up to 61.6% from June’s post-pandemic low of 61.5%. A soft payroll print alongside rising participation is genuinely weak. The same print with participation falling again means the low jobless rate is an artefact.
Average hourly earnings. Forecast at 3.5% year-over-year, unchanged. After Thursday’s unit labour costs at 1.3% and productivity at 1.4%, wage growth at or below 3.5% keeps the disinflation arithmetic intact. Above 3.6% and the Thursday relief partially unwinds.
Prior-month revisions. May was already revised down 43,000. A second consecutive downward revision would establish a deteriorating trend that a single headline cannot.
- Scenario Map
| Outcome | Threshold | Likely market reaction |
| Weak | Below 50k | Hike odds collapse well below 50%; front-end rallies; equities likely up but with growth concern. Watch participation to distinguish supply from demand. |
| Soft-consensus | 50–100k | The base case. Modest yield decline; September hike stays near a coin flip; equities supported. Most supportive outcome overall. |
| Strong | Above 150k | Vindicates the three dissenters; hike odds rise back toward 65–70%; Treasury yields higher and equities pressured — a strong print is now the bearish outcome for stocks. |
| Mixed | In-line headline, AHE above 3.6% | The awkward case: erases part of Thursday’s unit labour cost relief and pushes the long end higher. |
One dissenting framing worth holding. As one desk put it on Thursday, a September hike arguably should already be baked in given the strength of the second half of H1, and it is only at roughly 55% — a strong report could add to that while a weak one may not subtract much. The place to watch in the aftermath is the Treasury market, not equities.
And the out-of-consensus view: Citigroup calls for three rate cuts between now and January 2027, with the unemployment rate rising above 4.5% within months. If participation keeps falling and payrolls keep undershooting, that becomes the live scenario by the autumn.
- Carry-Over From Thursday
- Airbnb rose 7–9% after hours on an EPS and revenue beat with raised full-year guidance — and, notably, customer support cost per booking down 16% attributed to its AI assistant. Cloudflare jumped 16–17% on Q3 guidance above consensus.
- Datadog fell roughly 17% on a beat-and-raise, the fourth consecutive night a large winner has been sold on good news. The pattern to carry into Friday: exposure risk sits with the year’s best performers, not the worst.
- The SpaceX lockup is now live — up to 911.5 million shares against a float below 280.1 million, with further tranches through October, a second large release after Q3 earnings and the full backstop on 8 December. Friday is the first full session with unlocked supply.
- Crude rebounded on Hormuz scepticism, with the reported Oman proposal giving Tehran control of inbound shipping — the term Washington resists. Energy was the only sector higher on Thursday.
- The Road to September
| Date | Event | Why it matters |
| 12 Aug | July CPI | The most heavily weighted inflation input before the FOMC |
| 13 Aug | July PPI; SanDisk Investor Day | PPI completes the inflation picture; SanDisk addresses FY27 supply and price floors |
| 11 Aug | CoreWeave, Super Micro earnings | Late-cycle AI infrastructure read |
| 19 Aug | July FOMC minutes | Detail on the three-way dissent from Hammack, Kashkari and Logan |
| 27–29 Aug | Jackson Hole — Warsh’s first address as Chair | Framework and forward guidance after the Fed dropped explicit guidance |
| 15–16 Sept | FOMC decision | Currently ~55% priced for a 25bp hike |
Snapshot
Thursday, August 6th, 2026 — Theme: “Good Data, Red Screen” — Productivity beat by double, unit labour costs came in nearly a point below forecast and layoffs hit a two-year low. The Dow still fell 464 points and snapped a five-session record streak, as crude rebounded on Hormuz scepticism and yields rose into payrolls.
Thursday produced the most benign inflation arithmetic of 2026 and the market did not trade it. Q2 productivity at 1.4% against 0.6% expected, and unit labour costs at 1.3% against 2.1%, mean compensation growth of 3.4% is being absorbed by output rather than passed into prices — the first real evidence that AI capital spending is showing up in measured productivity. September hike odds duly fell to roughly 55% from 65%. But energy was the only sector to rise, Treasury yields moved higher, and positioning ahead of Friday’s payrolls dominated the tape.
🏛️ The Bottom Line
The Dow Jones Industrial Average fell 464.02 points (−0.85%) to 53,885, ending a five-day winning run of record closes, with Salesforce down 3% on a leadership shuffle a notable drag. The S&P 500 slipped 0.18% to 7,709.96 and the Nasdaq Composite eased 0.06% to 26,348.35. Energy was the only sector higher.
Thursday’s data was the week’s most important and the most constructive. Q2 nonfarm productivity rose 1.4% against a 0.6% consensus, with Q1 revised up to +0.8% from +0.3%. Unit labour costs rose just 1.3% against 2.1% expected, with Q1 revised down to 1.3% from 1.8%. Initial jobless claims came in at 199,000 versus 202,000–205,000 expected, with the four-week average falling to 198,750 from 203,250, though continuing claims rose to 1.801 million. Challenger job cuts totalled 33,429 in July — the fewest in two years, down 27% from June and 46% year-over-year.
Crude rebounded for the first time this week. WTI rose about 1.3% to $76.20 and Brent traded near $79.91–80.00 after Reuters reported the Oman-brokered proposal could hand Tehran control over inbound shipping through the Strait of Hormuz — a term Washington resists. Iran said the agreement was in its final stages while some Iranian sources rejected Trump’s claim that a deal was imminent, and Tehran is demanding the lifting of the US blockade of its ports. Gold traded around $4,267/oz (+0.5%); the dollar index was flat at 99.75.
Equity decliners were guidance casualties, not earnings misses. Datadog fell roughly 17% despite beating and raising, after nearly doubling in 2026; AppLovin −14% to −20%; Fiserv −12% on a cut to 2026 EPS guidance; Peloton −13%; Cedar Fair −18%; Solstice Advanced Materials −19%; SanDisk down a further 6–9%. Of 411 S&P 500 companies reported, 87% have beaten estimates — against 82% a year ago — with average year-over-year earnings growth of 25%.
Asia reversed Wednesday’s memory rally. The KOSPI fell roughly 2% with SK Hynix down more than 5%; the Nikkei dropped 617 points to 65,683 with Kioxia slumping 10% — a direct read-through from SanDisk’s guidance given their Japanese NAND joint venture. Shanghai rose 21 points to 3,900; Hang Seng fell 385 to 25,530. Europe held up: DAX +62 to 26,189, FTSE 100 +32 to 10,920.
After the bell, the non-crowded names delivered. Airbnb rose 7–9% to around $164.48 on EPS of $1.37 versus $1.25 and revenue of $3.61 billion versus $3.58 billion, raising full-year revenue and margin guidance — with customer support cost per booking down roughly 16%, attributed to its AI assistant. Cloudflare jumped 16–17% on Q3 guidance of $0.34 EPS on $736–737 million against $0.32 and $722 million. Instacart +8%; Lyft marginally higher; DraftKings −1.5%.
Two systemic warnings landed during the session. Jamie Dimon said market leverage is unusually high and that risks from borrowing outside traditional margin accounts are underestimated. Fed’s Musalem said inflation expectations remain anchored but conditions are fertile for them to unanchor, describing financial conditions as very accommodative with elevated asset prices.
📉 Reference Levels for the Friday Open (August 7th)
Derived from recent session closes and range extremes — not vendor-published levels. Verify against your own charts.
| Asset | Support | Resistance | Operational Bias |
| S&P 500 | 7,709 → 7,600 (Mon close) | 7,736 (Tue record) → 7,758 | 🟨 Range-bound into payrolls |
| Nasdaq Composite | 26,348 → 26,000 | 26,585 (Tue close) | 🟨 Holding better than the Dow |
| Dow Jones | 53,885 → 53,178 | 54,349 → 54,744 (records) | 🟥 Streak broken |
| US 10Y Yield | 4.55% → 4.50% | 4.69% → 4.73% | ⚠️ Payrolls is the trigger |
| US 2Y Yield | 4.15% | 4.27% | 🟨 Hike odds ~55% |
| Brent Crude | $78.77 → $75 | $83.77 → $90.12 | 🔄 Scepticism bid |
| WTI Crude | $74 → $72 | $80.00 → $84.67 | 🔄 Two-way |
| Gold | $4,200 → $4,100 | $4,267 → $4,350 | 🟩 Multi-week highs |
| KOSPI | 6,257 → 6,100 | 6,598 | 🟥 Guidance-driven |
| Nikkei 225 | 65,683 → 63,957 | 66,300 | 🟥 Kioxia drag |
📊 Market Sentiment & Bias
Macro: 🟩 The most constructive print of the year. Productivity at 1.4% with unit labour costs at 1.3% breaks the chain that turns 3.4% compensation growth into consumer inflation. Layoffs at a two-year low and claims below consensus remove the recession counter-argument. Caveat: preliminary productivity is heavily revised.
Equities: 🟨 Consolidation, not rejection. A 464-point Dow decline after five record closes, with the S&P down just 0.18%, reads as positioning into payrolls rather than a change of view.
Earnings regime: 🟥 Unchanged and unforgiving. Four consecutive nights of large winners sold on good news, now including a beat-and-raise from Datadog. With 87% of reporters beating, the beat is the base case.
Energy: 🔄 Credibility limit reached. The reported deal gives Tehran control of inbound shipping — the contested core of the five-month conflict. After $11 of Brent has been priced out, the market wants signature, not headlines.
Asia: 🟥 Now a guidance-transmission market. The leverage has been drained by Korean regulation; what remains is a one-session-lagged derivative of US after-hours guidance from SanDisk, Western Digital and AMD.
Systemic: ⚠️ Two credible warnings on the same day. Dimon on leverage, Musalem on accommodative conditions with elevated asset prices. Not timing signals, but both argue against maximum gross exposure at record index levels.
💡 Top Trade Takeaway: “The Data Improved; the Positioning Didn’t”
Focus: Continue rotating from the year’s largest winners into quality names where expectations have already reset. Favour the application and services layer over silicon. Treat energy as a binary policy trade. Keep gross exposure moderate through Friday’s payrolls, and watch the participation rate rather than the headline.
Logic. Thursday materially improved the macro backdrop and did nothing to change the microstructure problem. Unit labour costs at 1.3% with productivity at 1.4% is the arithmetic that lets the Fed stand still, and September hike odds fell to about 55% accordingly. But Datadog beat, raised, and fell 17% — the fourth consecutive night a doubled stock was sold on good news — while Airbnb and Cloudflare, neither of them crowded, rose 7–17% after hours. Better macro does not rescue crowded positioning; it just changes which names benefit.
The most important new information for the second half of the year is that AI is starting to appear on the return side of the ledger rather than only the spending side. Q2 productivity at 1.4% and Airbnb’s support cost per booking down 16% landed on the same day. If Q3 reporting extends that pattern, the framing shifts from “who can afford the capex” to “who is extracting the operating leverage” — which favours software, cloud and services over hardware. That is precisely the rotation the tape has run all week, and it now has a macro justification rather than only a positioning one.
The exposures to manage into Friday. Refining and energy remain priced on a closed Strait of Hormuz, and the contested term is who controls the waterway — this is a headline trade with a binary distribution. Asian memory exposure is effectively a leveraged position on US after-hours guidance, with SanDisk’s Investor Day on 13 August the next scheduled catalyst. And SpaceX now trades with up to 911.5 million unlocked shares against a float below 280.1 million, with further tranches through October and the full backstop on 8 December.
The honest caveat. One quarter of productivity data does not establish a trend, and preliminary readings are revised heavily — Q1 was itself revised from +0.3% to +0.8%. Core PCE remains at 3.3%, ISM Services prices hit 70.3 on Wednesday, and both Dimon and Musalem flagged elevated leverage and accommodative conditions on Thursday. The inflation case is weakened, not closed.
The report belongs to The Concept Trading and Van Hung Nguyen
