Retail Sales Printed 1.2% Against a 0.2% Call Hours Before the Fed Votes. The Hike Was Never the Question; the Dot Plot Is.

Wednesday, September 16, 2026, written at 9:20 a.m. Eastern, forty minutes before the open and four hours forty minutes before the decision. August retail sales landed at 8:30. Everything below is a pre-market mark and the entire picture can change at 2pm.

Top Call: The Consumer Just Contradicted Yesterday’s Survey, and the Risk Moved to the Dots

The Census Bureau published August retail sales at 8:30 this morning. Sales were $773.9 billion, up 1.2% on the month against a consensus near 0.2%, and up 6.0% on the year. The control group, which strips out autos, gasoline, building materials and restaurants and feeds straight into the consumption line of GDP, rose 0.7% on the month and 5.9% on the year. July was revised from down 0.6% to down 0.5%.

Read that against what we published twenty four hours ago. Yesterday the September Empire State survey collapsed from 20.6 to 7.6 with new orders falling to 2.0 and prices paid accelerating to 63.1, and we argued that it was the signature of an energy shock arriving in the real economy, that the committee was about to tighten into demand destruction already under way, and that the hike was therefore insurance bought against a cost the instrument cannot reach. One morning later the hard data says the opposite, and it says so with the cleanest series available.

We are not going to dress that up. The survey was a soft read on one region’s factories and the retail print is a national measure of what households actually spent. When those two disagree the hard number wins, and it won by a full percentage point against the call. Our framing was right about the mechanism and wrong about the timing, and anyone positioned for a Fed hiking into a visibly slowing consumer has to mark that.

The composition is where the useful argument is, so take it apart properly. The headline 6.0% year on year overstates volume, because gasoline stations are carrying a fuel price up roughly 27% over the same twelve months and every gallon bought at $4.33 counts as retail sales. Against August CPI of 3.4% the real headline is somewhere near 2.5%. But the control group excludes gasoline entirely, and it still ran 5.9%, which against a core rate of 2.4% is roughly three and a half points of real growth in the discretionary categories that the Federal Reserve can actually reach with a policy rate. That is the number that matters and it is not a consumer in trouble.

The honest caveat is the shape of the three month path. June was 0.3%, July was minus 0.5%, August was 1.2%. The average across those three is roughly a third of a percent a month, which is solid rather than spectacular, and part of August is a rebound from a July that Amazon’s June Prime Day had already hollowed out. One print does not make a boom. It does, however, comprehensively remove the case that the economy is buckling under the oil price, and that case was the only intellectually respectable reason to argue against tightening today.

So the hike itself is finished business. Futures have it at 92.5% for a move to a 4.00% upper bound, the first increase since July 2023, and there is no discount left in it. All of the risk sits in the Summary of Economic Projections released at the same moment and in the press conference thirty minutes later. A median dot above 4.00% puts a second hike live and makes the two year Treasury at 4.65% the wrong price, because the front end has anchored itself entirely on the assumption that today is one and done. A median at exactly 4.00% implies a pause, and in that case the long end keeps doing the work and the thirty year at 5.36% has no reason to stop.

Kevin Warsh does not submit a dot of his own, which he restated in June and which is consistent with a long stated view that the chart tells the market more than the committee knows. The practical consequence is that the median he does not vote in will be read as the committee’s path while the only person who can override it speaks at 2:30. That is an unusually wide distribution of outcomes for an event where the decision is 92.5% priced, and a VIX at 16.67 and falling 3.1% into it is not a market that has noticed.

The Credit Market Is Pricing SoftBank the Opposite Way From the Equity Market

Yesterday SoftBank rose 9.1% in Tokyo, and we wrote that the rebound was purchased with three announcements that the company and its affiliates intend to sell more securities rather than with any evidence that anyone is buying more compute. That was a financing argument and it now has a mark against it.

SoftBank’s five year credit default swaps are indicated near 384.6 basis points this morning, the widest since 2023, on concern about its OpenAI exposure. The SB Energy listing that produced yesterday’s equity move is targeting somewhere between $5 billion and $7 billion, OpenAI holds warrants in that same unit valued near $5.5 billion, and the parent has just retired the remaining $25.9 billion of its OpenAI bridge with a high yield sale of up to $20 billion planned behind it.

Two markets are looking at one balance sheet and reaching opposite conclusions in the same twenty four hours. The equity is pricing the listings as value crystallisation. The credit is pricing them as the reason the crystallisation is necessary. When those two disagree about a levered holding company, credit has historically been early and equity has historically been loud, and there is no version of this where 384 basis points is a bullish confirmation of a 9% rally.

This is the same point we have made for a week in a different currency. The buildout has moved from cash flow funding to capital markets funding, and capital markets funding reprices daily against a ten year Treasury that touched 5.04% yesterday. Bank of America guided third quarter investment banking fees to $1.6 billion to $1.8 billion against $2.0 billion on Monday and lost 5% for saying so. The arranger and the issuer are now both telling you the window is narrowing, and only one of them is being priced for it.

Asia Bought What New York Sold, for the Third Time in Four Sessions

The S&P 500 fell 0.45% on Tuesday to 7,585.73, the Dow lost 0.63% to 52,093.11 and the Nasdaq 100 gave up 0.65% to 28,937.84. Overnight the Kospi rose 1.4% to 6,717.97, the Nikkei added 0.7% to 63,923.00 helped by August exports up 19.3% against an 18.2% call, and the Hang Seng rose 0.2% to 24,713.78 and finally broke a five session losing run.

That is the third full reversal in the AI complex inside four sessions. Monday it fell apart on an essay about the pace of frontier capability. Tuesday it recovered on three financing announcements. This morning Seoul is up 1.4% led by memory ahead of a Federal Reserve meeting that has nothing to do with memory. None of those three moves was caused by an order, a shipment, a price increase or a cancellation.

A sector that round trips three percentage points in both directions inside a week on narrative and refinancing headlines is being set by positioning rather than by fundamentals, and the practical implication is that the next piece of hard evidence will move it more than any of this did. That evidence is dated. Micron reports on September 30 carrying more than $1 billion of HBM4 revenue already shipped, sixteen strategic customer agreements worth roughly $100 billion of minimum cumulative revenue and about $18 billion of customer cash deposits. Prepaid deposits are the least narrative contaminated demand signal in this entire theme, because nobody wires cash eighteen months early for capacity they expect to cancel.

The quieter structural story is the one The Information reported over the weekend and that nobody has priced. Palantir, Nvidia and Booz Allen are all restricting their internal use of frontier models from Anthropic and OpenAI pending irrevocable zero data retention guarantees, with Nvidia routing sensitive work to its own Nemotron models instead, and Microsoft is actively selling isolated cloud environments into exactly that anxiety. If enterprise adoption of third party frontier models is gated on contractual data guarantees rather than on capability, then the revenue ramp for the labs is slower and messier than the compute ramp, and the divergence between those two is the single most important unpriced variable in the sector.

Earnings: Trip.com Beats on Adjustments and Loses Money on the Mark

Trip.com reported after Tuesday’s close, and it is the only coverage relevant print in the window.

Our take: the operating business did what it was supposed to do. A 23% beat on adjusted earnings with revenue slightly ahead is a travel platform still taking share in a market where Monday’s data showed national retail sales growing just 0.4% year on year, and the gap between those two numbers is the point. Chinese households are not spending on goods and they are still spending on going somewhere, which is the same consumption substitution visible in most post-shock economies and a better signal for the Asian travel and hospitality chain than any official retail series.

The line that will get argued about is the $363 million GAAP loss against positive adjusted earnings, which is a fair value mark on the equity investment portfolio rather than an operating event. We would not pay for the loss and we would not ignore it either, because a company carrying enough listed equity to swing a quarter by more than a dollar a share has a second business it does not describe as one. Nothing here changes a constructive view on the operating asset.

Lennar reports around this afternoon’s decision with consensus at $1.30 on roughly $8.37 billion, earnings down about 35% on the year, against a thirty year mortgage in the mid sixes and a ten year that printed 5.04% yesterday. It is the cleanest available read on what a five handle does to a rate sensitive real economy, and it lands on the same afternoon as the dots. Treat the two as one event.

Overnight and Pre-Market Developments

• Skyworks: the best story in the index right now. Shares rose about 13% on Tuesday to near $89 and are indicated up another 9% to 13% this morning, taking the five day move to roughly 28% and the best weekly performance in the S&P 500. Two catalysts, both real. Apple’s iPhone Duo, the first foldable, was unveiled at $1,999 with an October 23 launch, and Skyworks is a core radio frequency supplier into that bill of materials at the highest price point Apple has ever charged. Separately chief executive Phil Brace told an investor conference that the $22 billion Qorvo merger is down to its final two regulatory jurisdictions. A content gain and a consolidation that removes the only credible competitor are a rare combination.
• Crypto: the Senate blocked the Clarity Act on a 50 to 49 cloture vote against the 60 required, stalling the market structure framework before the midterms. Bitcoin fell 3% and is at $75,681 this morning, Coinbase lost 8% and Circle 10%. The read is straightforward and unglamorous. Digital assets have been trading as a long duration risk asset priced off real yields for months, and removing the regulatory catalyst leaves them with nothing but a 5% ten year to trade against.
• Energy: crude gave back a little after Tuesday’s 4.4% surge in WTI to the highest settle since mid May, with WTI at $104.19 and Brent at $108.03. The giveback came on comments from Energy Secretary Chris Wright that the Saudi East-West pipeline will restart soon, and we would not pay much for that. The physical ledger says the opposite. No crude has left Yanbu since September 11 on Vortexa’s tracking, Aramco has told European customers that late September cargoes are cancelled, and cancelling cargoes is what an operator does when it does not expect the line back before the port runs dry.
• Oil inventories: the number nobody is quoting is the best argument on the page. Global observed inventories have fallen 507 million barrels since the war began, and August alone drew 95 million barrels, a rate of 3.1 million barrels a day. Official forecasts have a further 3.0 million a day of draw in the third quarter and 1.7 million in the fourth. A market drawing stock at that pace with Hormuz running five to seven transits a day against 130 to 140 before the war is not carrying an event premium. It is carrying a deficit.
• Ukraine and Russia: both sides struck energy infrastructure again, Russia hitting fuel stations in Kyiv and Ukraine hitting a Russian refinery, days after Washington announced a mutual halt on exactly those targets. Refining capacity is the tightest link in the global product chain right now and a failed truce over it belongs in the crude price.
• Gold: up roughly 0.7% to near $4,330 on spot with futures at $4,386.50, bouncing off Tuesday’s five week low near $4,290. We called that low the cleanest tell on the page yesterday. One bounce is not a refutation and the mechanism is unchanged, but the mark has gone the other way and we say so below.
• Europe and the United Kingdom: the Stoxx 600 rose 0.5%, the FTSE 100 0.4% and the CAC 40 0.3% with the DAX higher, recovering most of Tuesday’s slide. British unemployment held at 4.9% against a 5.0% forecast, which is a small hawkish surprise into Thursday’s Bank of England. German wholesale prices rose 6.8% on the year in August, which is the same energy pass through visible in every developed price index this month.
• Healthcare: the FDA granted accelerated approval to iberdomide with daratumumab and dexamethasone in multiple myeloma, and Vera Therapeutics reported that atacicept met all prespecified endpoints in the final ORIGIN 3 analysis in IgA nephropathy with two years of follow up. Clean data in a sector that has spent the month producing the opposite.

Key Events Today and This Week

Wednesday at 2pm and Friday in Tokyo remain one event rather than two. A hiking Federal Reserve and a hiking Bank of Japan inside three sessions removes the last cheap funding currency in the system during the same week that SoftBank’s credit is at a three year wide and its affiliates are queueing to issue. That sequence matters more than the twenty five basis points.

Geopolitical Risks

The Saudi East-West pipeline has now been shut for six days and the commercial consequence has arrived. Aramco has informed European buyers that late September cargoes are cancelled and Yanbu has loaded nothing since September 11. The line carries up to seven million barrels a day at capacity and is the only mechanism by which Saudi crude reaches water without transiting Hormuz. The American energy secretary says a restart is near. The cargo cancellations say the operator disagrees, and we weight the operator.

Hormuz itself is running five to seven transits a day against a pre-war 130 to 140, a collapse of roughly 95%. The Gulf Cooperation Council meeting with Iran in Oman was postponed after Saudi objections and a Bahraini refusal to attend, so the diplomatic track that briefly supported Friday’s selloff in crude no longer exists as a dated event. The Houthis hold Mokha, Perim and the Hanish islands, which is territorial control on both approaches to Bab el-Mandeb, and Libyan output has taken fresh setbacks.

Our framing is unchanged and the inventory data now supports it directly. The correct model is a structural transit and export constraint with an event premium layered on top, and only the second layer breathes. A market drawing three million barrels a day out of storage while its largest exporter cancels cargoes is not one where a verbal assurance about a repair timeline should take a dollar out of the price.

Elsewhere, Canadian tariffs on more than 700 American goods remain live with a federal procurement ban due September 29, and the meeting between Xi Jinping and the American president later this month still sits on unresolved rare earth and optical transceiver disputes.

Accountability

Two marks against us and one for us.

We wrote yesterday that Empire State was the most important number on the page after the yield itself and that the committee was tightening into a shock already destroying demand. August retail sales at 1.2% with a 5.9% control group says the national consumer did not get that memo. The mechanism we described is real and visible in the prices paid series, but we let a regional manufacturing survey stand in for an economy, and the economy answered within a day. That is a process error and it is recorded as one.

We also said gold falling to a five week low was the cleanest tell on the page, and gold is up 0.7% this morning. The argument that a nineteen year high in real yields outbids the haven still holds as a mechanism, and a bounce the morning of a dot plot is as likely to be positioning as conviction, but the mark has gone against us two sessions running on the volatility call and one on this. We are keeping both views and we are not pretending they are winning.

The one that is working is the funding argument. We said the AI buildout had migrated to capital markets funding and that the price of that funding was rising faster than equities were discounting. SoftBank’s five year protection at 384.6 basis points, the widest since 2023, arrived the morning after the equity rallied 9.1% on financing headlines. That is the thesis marking to market in the only place it can be observed cleanly.

Everything above is a pre-market mark taken at 9:20 a.m. Eastern. The cash open, the Saudi repair timeline and the median dot at 2pm can each rewrite the entire page.