The Dollar Made an 18-Month High in the Week Payrolls Missed by 60,000. That Trade Is Being Priced in Paris, Not Washington.
Monday, October 5, 2026  ·  Before Bell
Morning Note  ·  Equity Research

The Dollar Made an 18-Month High in the Week Payrolls Missed by 60,000. That Trade Is Being Priced in Paris, Not Washington.

October 5, 2026
Fund X
Steven Yuan
Dollar index 102.22 (+0.28%), an 18-month high
EUR/USD $1.1161 low, from 1.1255 Friday; 17-month low
France vs Germany 159bp 10-year spread, a cycle high
Spain Sanchez calls a snap election for November 29
October hike 17% to 22%; December near fully priced
Breadth Equal-weight S&P +0.33% vs index +0.73% Friday
Schneider / PTC $22.6bn cash at $205, a 42.3% premium
Foxconn Q3 revenue T$3.03tn (+47%); first T$1tn month
G7 reserves 100m barrels over four months, diesel first
Yemen Government offensive; Aramco Riyadh struck

Monday, October 5, 2026, written at 8:45 a.m. Eastern, forty five minutes before the open and seventy five minutes before ISM services. All futures and pre-market levels carry that timestamp and can move before the bell. Mainland China remains shut through October 7, and Korea is closed for a substitute National Foundation Day holiday. Hong Kong and Tokyo traded.

Top Call: A Dovish Week for the Fed Produced a Stronger Dollar, Because the Trade Is Relative and Europe Is the Weaker Side

Last week the American labour market printed 29,000 jobs against a consensus near 90,000, 60,000 was cut from July and August, and October hike odds fell from roughly 70% on Monday to between 17% and 22% this morning. Everything in that sentence is dollar negative. The dollar index is nonetheless at 102.22 this morning, an 18-month high, with the euro at $1.1161 for a 17-month low and the yen squeezed back above 158.

That is not a contradiction, it is a reminder that a currency is a spread and not a level. The euro is being sold because France’s ten year spread to Germany reached 159 basis points, a fresh cycle high and the widest since the sovereign crisis of 2011, with French public debt projected at 121.7% of GDP in 2027 and Lecornu attempting 54 billion euros of savings through a parliament that has already removed one prime minister over a budget. On Sunday Spain’s Sanchez called a snap election for November 29. The dollar is winning a contest it is playing badly.

The consequence is the part worth holding. A dovish Federal Reserve is supposed to deliver easier global financial conditions through a weaker dollar. This time it has not. A softer Fed plus a stronger dollar tightens conditions for everyone outside the United States, and the economies it tightens for are precisely the ones importing Brent at $101 with no domestic offset. Europe gets the energy bill, the currency, and now a sovereign spread, all at once. That is the mechanism by which an American labour report becomes a European problem, and it is why we read this morning’s euro print as a more important market event than anything on the domestic calendar before 10:00.

Rates: The Front End Kept Its Payroll Rally and the Long End Gave It All Back

The two year sits near 4.80% having dipped to 4.69%, against 4.839% at Friday’s close. The ten year is 5.26% to 5.30%, which is where Friday left it after a session that opened at 5.176% four minutes past the release and closed five basis points higher at 5.281%. The thirty year is near 5.62%.

So the round trip we flagged on Saturday is now a settled pattern rather than a single afternoon. The front end holds a policy repricing. The long end refuses to participate. Two year to ten year is wide near 46 to 50 basis points, and the thirty year at 5.62% is still a July 2002 level on the most dovish labour print of the cycle.

The market reads that as a steepener and leaves it there. We read it as the bond market pricing a committee that stops before the inflation does. October is now 17% to 22% and December is close to fully priced, which is the same hike moved by eight weeks rather than a hike removed. For anyone funding at the long end, a December move is about as restrictive as an October one, and the term premium required to hold thirty year paper against this deficit and this supply calendar does not care which meeting it lands in. Wednesday’s minutes from the September 15 and 16 meeting will be read for how many members carried the one more hike guidance that Williams repeated on Friday. We expect that count to be higher than the current pricing implies, and we note that minutes three weeks stale against four data points are a weak instrument for resolving it.

Gold is roughly flat between $4,140 and $4,187 and bitcoin is up about 1% near $86,100, with spot bitcoin funds having taken $2.65bn in September for their second largest month since October 2025. Nominal yields are lower at the front end and gold did not rally, which is the first clean failure of the carry framework as we have been publishing it, and the reason is on the screen above it.

Energy: Brent Is Back Over $100 on Yemen, and the G7 Has Conceded the Argument on Diesel

Brent is near $101.40, down about 0.8% on the session but back above the hundred handle after settling $99.68 on Friday and falling roughly 4.4% on the week. WTI is near $90 on the November contract.

The weekend supplied the bid. Yemen’s Saudi-backed government launched a full offensive to retake territory from the Houthis, who had captured the Bab el-Mandeb strait and some 150 kilometres of Red Sea coast in the preceding weeks. The Houthis responded by claiming ballistic missile and drone strikes on a Saudi Aramco facility in Riyadh on Saturday, with intercept debris landing in Asir. A third American carrier strike group has deployed. Trump said Iran will do this the easy way or the hard way and declined to say which. OPEC+ left November targets unchanged on Saturday, a decision with almost no information in it while core output of 25.0m barrels a day sits around 5m below prewar levels and the spare capacity that exists is sitting in the two countries being shot at.

Two developments cut the other way and both deserve more attention than the headline. Iraq’s state tanker company has chartered a VLCC to move 2m barrels past Hormuz, described as its first long haul commercial voyage beyond the strait in decades, which is a seller finding a route rather than a diplomat finding an agreement. And the G7 has agreed to release up to 100m barrels of crude and fuel over four months, roughly 833,000 barrels a day, or about one sixth of the Gulf shortfall.

The composition of that release is the news. It is front-loaded with diesel. For six weeks this desk has argued that the shortage lives in refined product rather than crude, that the diesel crack above $100 and distillate inventories at the lowest September level since 1982 were the real signal, and that a market reading falling WTI as energy disinflation was reading the wrong barrel. The policy apparatus has now written that down. American retail diesel is near a record $6.50 a gallon and European diesel futures have traded above $200 a barrel. A reserve release aimed at product rather than crude is an official admission that refining capacity, not wellhead supply, is the binding constraint. It is also, at one sixth of the gap, a price signal rather than a fix.

Equities: A Record Index With a Median Stock That Barely Moved

Friday closed with the S&P at 7,722.72 up 0.73%, the Dow at 51,176.96 up 0.49%, the Nasdaq Composite at a record 27,190.86 up 1.19% and the Nasdaq 100 at a record 30,807.93. The SOX added 2.4%. The equal-weight S&P rose 0.33%.

That last number is the session. The cap-weighted index did better than twice the median stock on a day the semiconductor complex carried everything, and the index printed a record while most of its constituents went roughly nowhere. We spent late September arguing that the index had become one trade, and on Friday morning we logged the first counterexample when the Russell contract was the strongest of the four. The cash close partly retires that counterexample, with one honest amendment attached: the Russell did close up 0.94%, ahead of equal weight. Breadth is weak in the middle of the large cap index rather than at the bottom of the market, which is a different claim from the one we made and a less convenient one.

Futures are slightly lower this morning, with S&P contracts near 7,770, Nasdaq contracts down about 0.2% and the Russell the only one positive. The VIX is 16.19, up almost 6% from Friday’s 15.31, which is the Hormuz and Yemen premium returning rather than anything domestic.

Asia and Europe: Tokyo Touches 70,000, Hong Kong Steadies, Paris Cannot

The Nikkei closed near 69,947, up about 2.4% and roughly 1,640 points, after touching 70,037.61 intraday for the first time in about three months. The driver is the same global discount rate the Americans priced on Friday, amplified by a yen at 158.33 against the dollar, a one year low, which flatters every exporter in the index. Tokyo is now trading a currency that is weakening for reasons that have nothing to do with Japan, which is a good quarter for the earnings translation and a bad one for the import bill, and the Bank of Japan has to answer for the second part eventually.

Hong Kong recovered about 0.3% to roughly 24,040 after Friday’s 2.60% drop, its largest single session decline since March, taken on the first session back from the National Day break. The mainland is absent until October 8 and has therefore missed the entire data week. Korea is shut today for a substitute holiday with the Kospi last at 7,003.74, into a month in which foreign investors have now sold something close to 20 trillion won of Korean equity.

Europe is mixed and the dispersion is the point. The DAX is near 25,194 and slightly lower, the FTSE is firmer near 10,486, and the CAC is down about 0.6% with French budget risk the named cause. The French ten year reached 4.995% last week before closing at 4.87%. A market that was willing three weeks ago to treat French politics as noise is now pricing it as credit, and it is doing so while the ECB remains a non-factor in the conversation. European equities have spent a month trading the American discount rate instead of their own price level and their own sovereign curve. That works until the sovereign curve forces the issue.

Corporate: Schneider Pays $22.6bn in Cash for PTC, and the Buyer Is French

Schneider Electric agreed to acquire PTC for $205 a share in cash, valuing it at $22.6bn, a 42.3% premium, unanimously approved by both boards and targeted to close by the third quarter of 2027. PTC rose about 36%. It is Schneider’s largest deal ever, roughly double the $11bn Aveva purchase in 2023, and it is underwritten by about 800m euros of revenue synergy by year three against 250m euros of cost savings.

Three things make this more interesting than a software premium. First, Schneider is an electrical equipment manufacturer buying CAD, product lifecycle management and industrial internet of things software. The thesis is that the scarce asset in industrial AI is not the model but the design and lifecycle data that describes physical systems, and that the company which owns the electrical infrastructure should also own the digital twin running on it. Second, read it against third quarter global M&A of $993bn, down 41% and the first sub-trillion quarter since the second quarter of 2025. The deals getting done in a closed window are strategic buyers paying cash for assets adjacent to things they already operate, not financial sponsors levering multiples. Third, and least discussed, a French acquirer is committing $22.6bn of cash on the morning the euro prints a 17-month low, into a funding market where French sovereign spreads are at 2011 wides. The currency alone has made this roughly three percent more expensive than it would have been a week ago. If European corporates are going to keep buying American software, the fiscal story in Paris is now a line item in their deal models.

Foxconn reported third quarter revenue of T$3.03 trillion, about $95.4bn, up 47% year on year, including a record September above T$1 trillion for the first month in its history, roughly $36.5bn. AI server revenue passed half of total revenue for the first time. This is the cleanest real time read on Nvidia rack shipments available to anyone, and it says the physical buildout is still accelerating rather than digesting.

The qualification is in the margin line. Foxconn’s revenue grew 47% while profitability slipped, because AI server assembly is a larger, lower margin business than consumer electronics assembly. A company whose top line compounds at that rate while its earnings do not is telling you exactly where the economics sit in this stack, and the answer is not with the integrator. That is consistent with the memory and accelerator margins we have spent two weeks discussing, and it is the reason we keep separating the volume signal from the profit signal in this complex.

Metric Prior year Actual Read
Q3 revenue About T$2.06tn T$3.03tn (+47%) Volume signal, not a margin signal
September revenue Below T$1tn Record, above T$1tn First trillion-NT-dollar month ever
AI server mix Under half Over half of revenue Mix shift is accretive to sales, dilutive to margin
Profitability Higher Slipping Economics sit above the integrator in the stack

Cerebras is the more important technology story and it round-tripped in five sessions. The stock fell 19.5% last week to its lowest level since the May listing after SemiAnalysis reported that OpenAI’s new flagship, GPT-6.1 Sol, is serving its Ultrafast tier at roughly 300 tokens a second on conventional Nvidia GPUs running at low batch size, rather than on Cerebras wafer scale parts. It rallied this morning after Sam Altman called the company a close partner.

Strip the sentiment out and the technical claim is the whole investment case for specialised inference silicon. The argument for wafer scale has always been that GPUs cannot serve low latency, low batch decoding at an acceptable cost, which is why a token-per-second tier should need different hardware. If Nvidia parts can do it at low batch economics, that moat is narrower than the price assumed, and Cerebras has roughly $10bn of OpenAI capacity to deliver by 2028 against it. A chief executive describing you as a close partner is not a purchase order, and this morning’s bounce is a sentiment trade on a structural question that has not been answered.

Vistra is receiving a $4.2bn federal loan, to be announced today by Energy Secretary Wright at its Lake Erie site, to fund power uprates at Perry and Davis-Besse in Ohio and Beaver Valley in Pennsylvania. Uprates on existing licensed reactors add firm baseload inside PJM, the tightest capacity market in the country, without requiring new NRC licences. This is the cheapest marginal electron in the entire artificial intelligence buildout, and it is worth noting that the federal balance sheet is now financing the power leg of this trade directly, which is a materially different intervention from financing the silicon.

Qualcomm signed a multi-year patent licence with Huawei covering 5G, AI, computing and networking, Huawei’s first licensing agreement with Qualcomm to cover 5G. It has two legs: a cross-licence of both portfolios, and Qualcomm purchasing certain Huawei United States patents in compute, AI and networking, subject to approval. Huawei says the deal takes its total licensing book above $6.9bn. Qualcomm is licensing patents underpinning Huawei’s LogicFolding chipmaking technique. Huawei has become a licensor to Qualcomm rather than only a licensee, and American silicon is paying for Chinese process intellectual property in the same month Washington moved against Chinese optical transceivers. The export control architecture is pointed at hardware and the value is moving through patents.

Elsewhere, Cenovus is acquiring Athabasca Oil at a C$5.7bn enterprise value, Intel fell 3.29% on Friday and is under pressure again after reports that Musk is in talks with TSMC for his Terafab venture, which undercuts the foundry narrative at the exact moment Intel needs an anchor customer, and Vaxcyte reported that VAX-31 met its late stage endpoints with results comparable to the Pfizer and Merck pneumococcal franchises.

Key Events Today and This Week

When What Why It Matters
Mon 10:00 ISM services, consensus 55.0 The prices component, not the headline
Mon 13:05 Williams, then Bowman at 14:45 He carried one more hike guidance on Friday
Wed 14:00 FOMC minutes, September 15-16 How many members carried that guidance
Wed, close Applied Digital, $135m and -$0.30 The only AI data centre print of the week
Thu 08:00 PepsiCo, $25.0bn and $2.30 First read on staples pricing power
Thu TSMC September revenue August was +53.3%; the cleanest AI tape
Fri 08:00 Delta, $17.59bn and $2.03 Jet fuel against a $100 distillate crack
Oct 8 China and Hong Kong fully return Absent for the entire data week
Oct 27-28 FOMC Priced 17% to 22%; December near 100%

ISM services at 10:00 is the line we named on Friday and it is the one to read today. Manufacturing told us on Thursday that goods inflation is accelerating at the factory gate, with prices paid at 77.9 and not a single commodity category reporting a decrease. The services survey covers roughly three quarters of the economy.

We will commit to both readings now so that neither can be selected afterwards. If the services prices index follows manufacturing higher while the headline holds near 55, then the committee is being asked to pause into accelerating inflation and decelerating employment simultaneously, which is the one configuration it has no good answer for, and the October pause becomes a policy error in progress rather than a reasonable call. If services prices fall while manufacturing’s rose, then last Thursday’s 77.9 was a goods-specific pass through from tariffs and energy rather than an economy-wide price impulse, and we will have overweighted a single survey for four sessions. We are not predicting which. We are writing down in advance what each would mean.

Accountability

The carry framework fails cleanly for the first time and the reason is a term we left out. We have published gold and bitcoin as a function of real yields and restored it to full confidence on Friday after it worked in both directions. This morning front end nominal yields are lower than Friday’s close and gold is flat. The framework as written predicted gold higher. It is not wrong so much as underspecified, because it has no dollar term in it, and the dollar is at an 18-month high. We are adding that term rather than claiming the model survived, and we should have had it in from the beginning given we spend most of our time on cross-border flows.

The breadth call is restored and narrowed in the same session. Friday’s equal-weight S&P rose 0.33% against the index at 0.73% with the Nasdaq 100 at a record, which is the one-trade index we described in late September. But the Russell closed up 0.94%, ahead of equal weight, so the counterexample we logged on Friday morning is not fully retired. The accurate statement is that breadth is weak in the middle of the large cap index, not across the whole market, and that is a narrower claim than the one we made.

The distillate thesis gets its strongest confirmation in six weeks, from a direction we did not name. We argued the shortage sits in refined product rather than crude and that falling WTI was the wrong barrel to read. The G7 has now announced a 100m barrel release front-loaded with diesel, which is the policy apparatus writing our argument down. We take the structure and not the timing: for three of those six weeks we named the wrong trigger, and the confirmation arrived through an emergency reserve decision rather than through any of the market mechanisms we pointed at.

The crude discipline rule pays for an eighth week. Brent fell about 4.4% last week on inventories and demand downgrades and is back above $100 three sessions later on a Yemen ground offensive that was not in anyone’s forecast on Thursday. Carrying no directional view on the barrel while trading the product crack remains the right posture, and we have now been paid for it through two full round trips.

On timestamps. Saturday’s edition marked a timestamp error on the two year, where we published a four minute post-release level against a close. Every rate, future and currency level in this note carries an 8:45 a.m. Eastern stamp, and where we compare to Friday we say whether we mean Friday’s close or Friday’s morning. That is a process fix rather than a view, and it is the kind of error that is cheap to prevent and expensive to repeat.