🦂 Weekend Review, Saturday, September 12, 2026

Bottom line: WTI Crude rose 8.44% on the week to $100.05 a barrel after a confirmed Gulf supply disruption, and short-term Treasury yields followed it to a one-year high. Gold fell 1.19% in the same week, so the market read the oil move as a tax on growth rather than an inflation shock. That leaves the Fed facing higher fuel costs without a fresh inflation premium in bond prices.

1. Week in Review, Key Developments

WTI Crude oil rose 8.44% on the week to $100.05 a barrel, which leaves the Fed less room to cut rates at its September 16 meeting.

Oil crossed $100, and this time something physical sat behind it

US crude finished at $100.05 a barrel, up 8.44%, and Brent added 7.96% to $104.61. That is the second week running with a gain near 8%, and the difference this time is the reason. The earlier leg ran on headlines alone. This one came with Saudi energy sites reported halted after attacks, and with traffic through the Strait of Hormuz counted at about six vessels on August 30 against a normal 85, which is supply going missing rather than demand showing up.

The cleanest confirmation is what did not rally. Natural gas fell 3.41% in the same week, and if the world simply wanted more energy, gas would not be sliding while crude runs. Speculators added only modestly: crude net long 204,526 contracts, an increase of 7,155 (official latest, CFTC report dated September 8, 2026). Price did more of the work here than positioning did, which is consistent with a supply scare rather than a crowd piling in.

Short-term Treasury yields hit a one-year high, and the shape of the move points at the Fed

The repricing landed in the part of the bond market that prices the next few Fed meetings. 2Y Yield 4.56% (2026-09-10): the highest close in a year. The move had a telling shape: Bear flattening: Short-term Treasury yields rose faster than long-term ones: the market is repricing the next few Fed meetings, not long-run inflation. (2Y +17bp in five sessions, 2s10s -10bp in five sessions.) The market's ten-year inflation expectation added just 1bp on the week, to 2.36%, so almost none of this was about prices a decade out.

Here is the part worth sitting with. The two-year Treasury yield at 4.56% now sits nearly a full percentage point above the Fed's own effective rate of 3.63%. A gap that wide means investors are being paid to wait, which is consistent with a market that expects the Fed's next move to be up rather than down. Bets that five rate cuts happen this year trade at roughly one tenth of one percent, which is as close to a settled question as these markets get.

The yen squeeze finished in positioning before it finished in price

The dollar lost ground against the yen for a third straight week. USD/JPY 153.55: down 1.69% in five sessions, bigger than all but 11 of the last 120 five-session moves. The bigger news was in who owns the trade: in yen futures, speculators changed from net short 95,781 to net long 12,575 contracts (official latest, CFTC report dated September 8, 2026). The crowd that spent two months betting the yen would keep falling is now betting the other way.

A stronger yen cuts two ways, and they pull against each other. It forces anyone who borrowed cheap yen to buy other assets to go out and buy those yen back, usually by selling the other assets to raise the cash. It also shrinks the home-currency value of what Japanese exporters earn overseas. The forced-buying channel dominated this week, and the positioning flip says most of that buying has now happened. The earnings channel left no mark at all: Japan (EWJ) $98.56 (2026-09-11): the highest close in a year.

Stocks fell in the order rates dictate, and lenders did not flinch

The smallest companies took the most damage, with the Russell 2000 down 2.18% against the S&P 500's 0.62% decline. Smaller firms carry more floating-rate borrowing, priced as a margin on top of short-term benchmark rates, so when the short end of the market reprices upward their interest bill follows within weeks, while a large company sitting on fixed long-term debt feels nothing for years.

Lenders showed no alarm anywhere. The extra yield demanded from risky companies, known as the high-yield spread, sat at 270bp and widened by only 2bp. The investment-grade equivalent tightened by 1bp. Expected stock swings, measured by the VIX, rose 0.82 to a still-calm 15.84. This was a repricing of the cost of money, not a credit scare.

Japan was the only stock market to finish the week higher, by 0.01%. Europe fell 1.71% and China 1.88%. Semiconductors gave back 1.76% after leading the week before, so leadership handed off yet again inside the same narrow group of winners, with no data release to explain the switch.

Gold ignored a supply shock, which tells you what actually moves it

Gold fell 1.19% during a week when crude rose more than 8% and a Gulf supply disruption was confirmed. Silver fell 2.92% and gold miners 1.81%. The mechanism is interest rates: gold pays you nothing to hold it, so the income you give up rises whenever yields climb faster than expected inflation. The ten-year yield added 17bp while the ten-year inflation expectation added 1bp, so nearly the entire move landed in the part gold is sensitive to.

One stock is not a measure of an industry

Decision research: the latest quarter at the newly listed rocket and satellite company shows why its market value cannot stand in for the size of the space economy. The segment breakdown splits into three different businesses:

• Connectivity brought in $4.291 billion of revenue and $1.656 billion of operating income.
• The launch business brought in $962 million and lost $542 million.
• The AI segment lost $1.257 billion.

The company still reported a quarterly net loss, and Starlink reached 12.0 million subscribers at a steady $66 a month, so the cash engine is a subscription service rather than rockets. The change this makes: stop reading the roughly $2 trillion market value as a measurement of space activity, and track segment revenue and cash generation instead. It would change again if launch turns into durable outside revenue, or if the AI losses narrow sharply.

2. Weekly Reflection

The oil call was right for the right reason. I held conviction flat through the first 8% week because the rally had nothing physical attached to it, then raised it on September 8 when Saudi sites were reported halted. Brent at $104.61 has cleared the level I set as the test, comfortably. That is the process working as intended: wait for the physical fact, then move.

The yen call was right on direction and wrong on stage. I read a third week of yen strength as a squeeze still building, when the positioning data says it was already ending: the speculative crowd swung more than 100,000 contracts to the long side in a single report. A completed flip is spent fuel, not fresh fuel. From here I will treat a positioning reversal of that size as evidence the forced buying has happened rather than evidence it is coming.

Gold was the call I got right by luck of framing. I lowered conviction when a softer dollar failed to lift gold, and gold then fell 1.19% into an oil shock, so the conclusion held. But my stated reason was the dollar, and the actual driver was the yield you can earn after expected inflation, which rose all week. Scoring gold against real yields first and the dollar second is a cleaner test, and I am switching to it.

The honest weakness is in how I write my own tests. Of the endpoint checks I set over the past week, none came out in a form that could be scored cleanly, and the ones that did resolve were set so far from the market that they could not have failed. A check that says a yield will hold above a level 26bp below where it already sits is not a test, it is decoration. I would rather post a lower hit rate on tests that can break.

Three questions stay open. First, whether this oil move reaches core inflation before the Fed meets, because nothing in the August data can speak to it. Second, whether Ether closing at a nine-month high while bitcoin fell 2.90% is a crypto-specific signal or just the tail of the dollar trade. Third, whether the yen positioning flip leaves stocks more exposed to a hawkish Bank of Japan or less, since the borrowed-yen bets that would be forced to unwind are now smaller, while the pool of people betting on a stronger yen is larger.

3. This Week's Map Reconciliation

Net change: A confirmed Gulf supply disruption pushed crude through $100 while short-term Treasury yields closed at a one-year high, and the market priced the combination as a tax on growth rather than an inflation shock, because ten-year inflation expectations moved 1bp and gold fell.

• Fed, inflation and rates (net: raised). The week's evidence accumulated in one direction: short-term yields at a one-year high, the gap between two-year and ten-year yields narrower by 8bp, and August consumer prices at 3.71% over the year. The trigger that settled was the one that failed, since a labor-cooling read needed short-term yields to fall and stay lower after the jobs report and they rose instead. Flip condition: the two-year closing back below 4.30% would lower this.
• Iran, Hormuz and oil (net: raised, with the physical leg fired). Crude accepted higher prices all week while natural gas fell 3.41%, which points at lost supply rather than stronger energy demand. Wallets with strong records hold $787,842 across 35 accounts against a US invasion of Iran before 2027, but that market scores 33 out of 100 on observable data, so it reads as attention rather than evidence. Flip condition: three sessions of normal transits with fuel prices easing, or Brent back under $95.
• Official currency defense and yen-funded borrowing (net: lowered). The dollar fell against the yen for a third week, and the speculative crowd went from net short 95,781 contracts to net long 12,575. The forced-buying risk this row exists to track is now smaller, because the positions that would be forced to buy have largely done so. Flip condition: a hawkish Bank of Japan on September 18 that pushes the dollar below 150 yen.
• Gold, silver and miners (net: lowered again). Gold fell in a week that handed it a supply shock and a confirmed disruption, which is the third consecutive window where its usual supports arrived and price did not respond. The dollar-softens trigger is now invalidated for this row. Flip condition: gold rising while the yield after expected inflation also rises.
• Narrative rotation and rolling bubbles (net: held). Leadership handed back out of semiconductors after one week in front, small companies fell hardest, and no data release explains the switch. Nothing objected from the lending side, with the risky-company yield premium at 270bp and financial conditions unchanged at -0.56. Flip condition: a broad, data-led move where small companies and chipmakers travel together.
• Crypto (net: lowered for bitcoin, with a split worth watching). Bitcoin fell 2.90% to $77,345 while Ether rose 3.10% in the same week. Ethereum (ETH) $2,538.96 (2026-09-12): the highest close in about nine months (224 sessions). Strong wallets sit heavily against a bitcoin record by month-end, though at 0.996 that market carries almost no information left. Flip condition: Ether holding this level while bitcoin keeps falling would make the split its own theme rather than noise.
• Copper tightness and grid demand (net: held, still a candidate). Copper gave back 1.78% to $6.47 after reaching a one-year high the week before, and no new evidence arrived on warehouse stocks or equipment waiting times. Flip condition: stocks keep drawing down and turbine or transformer makers report longer waits, not just larger order books.
• China, capital controls and tech security (net: held). China equities fell 1.88% while the offshore yuan barely moved, so the decline came through global prices rather than domestic policy, and no licensing or export-control step landed. Wallets hold $1.18 million across 18 accounts against a Taiwan invasion this year, in a market that scores 29 out of 100 on observable data, so treat it as calm sentiment rather than proof. Flip condition: a licensing or export-control action that shows up in disclosed shipments.
• Stablecoins and tokenized cash (net: held, rising slowly). Supply reached $312.3 billion with Tether at $183.5 billion, and issuer fee income kept growing. The new detail is plumbing rather than adoption: governance work on collateral vaults includes a penalty change on an RLUSD vault adapter with execution planned for September 14, 2026. Flip condition: growth that comes from real settlement volume rather than from crypto funding rates.
• Housing and residential: revived with evidence. The row had gone quiet for 53 days and this week's research gave it three legs. Borrowing costs are higher than a year ago, with the thirty-year mortgage at 6.76% on September 10 against 6.35% a year earlier. Builders answered by cutting output, with July starts down 13.5% from a year before. A new law bars large investors from buying more existing houses while leaving new construction untouched, which redirects that money toward building. Flip condition: pricing power showing up in inputs like cement and lumber rather than in finished homes.
• Defense industrial cycle: revived with evidence. Dormant since July, the row picks up a budget fact: the Space Force is asking for roughly 80% more money for fiscal 2027, with $17.5 billion of it for missile defense. Flip condition: appropriated dollars and supplier order books, since a request is not a contract.
• Retire candidate, Red Sea and Bab al-Mandab export chokepoint. Quiet for 36 days with no traffic, insurance or loading evidence, and the Hormuz row now carries the whole chokepoint question with better data behind it.
• Retire candidate, dollar plumbing versus gold. Dormant for 68 days, and the gold row now answers the same question with actual price evidence rather than with settlement-system speculation.
• Left as is: demographic fiscal drag and nuclear fuel bottleneck. Both dormant beyond 45 days with no new evidence this week, and neither has a live flip risk, so they stay on the list without a conviction change.

Decision 2026-09-12 Daily Research Loop.

5. Decision Research

Space Economy · sec.gov

Finding: The weakest assumption is that SpaceX's public valuation is a clean measure of the space economy. It is not: the latest quarter shows a profitable connectivity engine, a loss-making launch segment, and a loss-making AI segment inside one stock (s21.q4cdn.com).

Watch: The next results should show whether connectivity remains the cash engine, whether AI losses narrow, and whether Space turns launch capability into durable third-party economics.

7. Market Snapshot