Oil Goes to $99 as Houthis Hit Aramco Sites and Saudi Arabia Halts Operations. Canada's Retaliatory Tariffs Go Live.

Tuesday, September 8, 2026, written at 9:00 a.m. Eastern, half an hour before the first honest American mark since Friday’s close. Monday was Labor Day and the cash tape was shut, so everything below has been priced by Asia, by Europe and by a thin futures session, and the 9:30 open is the first time US capital gets a vote on any of it.

Top Call: The War Moved From Ships to Wells

For six weeks this conflict has been a transit problem. Tankers, war risk premiums, a corridor running six to eight passages a day against a normal eighty five. This morning it became a production problem. Houthi ballistic missiles and drones struck Aramco facilities at Jizan, Abha and Najran along with King Khalid Air Base at Khamis Mushait, wounding 73 people including women and children, and the Saudi energy ministry confirmed that operations at some facilities have been halted. Brent went to a seven-week high between $98.62 and $99.73 and WTI to roughly $93.

The reason this is different from every previous escalation is arithmetic rather than emotion. A blocked strait delays barrels; a damaged facility deletes them. And it lands forty eight hours after OPEC+ met and left quotas flat for the rest of 2026, having already completed the rollback of the entire 1.65 million barrel voluntary cut. The organisation whose function is to absorb exactly this kind of shock spent Sunday confirming that it has nothing left to absorb it with. We would treat $100 Brent as the reference case into the September 16 meeting rather than as a tail, and we think energy is now an inflation input for the fourth quarter rather than a headline risk.

The Geography Is Worse Than the Damage

The damage assessment is not in yet and Aramco has not commented, so nobody should be extrapolating a barrel count from a fire. What can be assessed immediately is where the fires are.

Jizan, Abha and Najran sit on and behind Saudi Arabia’s southwestern Red Sea flank, not on the Gulf coast. That matters because the entire institutional answer to a Hormuz closure has always been the East-West Pipeline, which carries crude across the kingdom to Yanbu on the Red Sea and out through Bab el-Mandeb, bypassing the strait completely. That bypass runs down the same coastline the Houthis can now reach with ballistic missiles and drones, and Houthi military spokesman Yahya Saree used this morning’s claim to declare a broad campaign against Saudi oil infrastructure rather than a single retaliation.

So the market is no longer pricing one chokepoint with a workaround. It is pricing a chokepoint and its workaround under simultaneous threat, from two different actors, with the world’s swing producer having formally stood down its own spare capacity two days ago. That is a structurally different distribution from the one that governed crude in July and August, and it is why a move of only about one and a half percent this morning strikes us as an underreaction rather than an overreaction.

The one genuine offset is diplomatic and it is close. Iran says a deal with Oman on managing Hormuz shipping is days away, including maps and a temporary safe corridor, alongside the maritime exclusion zone Tehran intends to declare outside the strait. That is an attempt to convert a stalemate into a toll booth, and it is bearish for crude in the narrow sense that transits would resume. It is bullish for Iranian leverage in every other sense, because a corridor operated by permission is a corridor that can be closed by permission. Watch whether Washington accepts it.

The Futures Board Is Telling You Who Pays

Dow futures are down 344 to 389 points, roughly 0.64% to 0.78%. S&P futures are down about a quarter of a percent. Nasdaq futures are green. That dispersion is not noise, it is the entire morning compressed into three numbers.

Canada’s retaliation took effect this morning: duties of 15% to 50% on roughly $27.6 billion of American goods, with steel, aluminium, iron, furniture and clothing taking the full 50%, plus dairy, agricultural equipment, paper, appliances and electronics. Ottawa has layered a $7.5 billion support package on top of the $25 billion committed since April 2025. Add a crude price near $100 and you have a tax on precisely the industrial, transport and consumer complex that dominates the Dow, and almost no direct claim at all on the semiconductor and software complex that dominates the Nasdaq 100.

This is the same layered rotation we have been describing for two weeks, arriving through a completely different door. Last week capital moved down the stack because a ten year at 4.80% marks down terminal value while sold out memory capacity gets marked up on cash flows arriving inside eighteen months. This morning it is holding position because tariffs and oil are input costs for things that move atoms and are close to irrelevant for things that sell allocation of a scarce chip. SanDisk is bid pre-market on reports that large cloud and AI customers are locking multi-year NAND supply as far out as 2031, against $42 billion of contract revenue already booked, more than a third of fiscal 2027 output pre-sold, and no new wafer capacity landing before 2027 or 2028 while demand compounds above 20%. Goldman’s $7.6 trillion of AI infrastructure spending through 2031 is the demand side of that same sentence.

The confirmation that this has become structural rather than thematic is quieter and more revealing: effective September 21, Dell, Palo Alto, Arista and SanDisk enter the S&P 100, displacing Honeywell Aerospace, Nike, Simon Property and Colgate-Palmolive. Index committees are slow and mechanical. When they act, the reallocation has already happened.

Japan Is the Risk Nobody Has Priced

The Nikkei fell 1,130.51 points to 65,269.29, down 1.70%, giving back roughly four fifths of Monday’s 2.12% memory-led rally inside one session. The cause was not chips. It was the currency. The yen strengthened as much as 1% to 152.89 per dollar after gaining 1.2% on Monday, reaching a seven-month high, on hardening expectations that the Bank of Japan hikes at its September 18 meeting.

Here is the part that deserves attention beyond Tokyo. The Federal Reserve decides on September 16 with 58% to 66% priced for a hike. The Bank of Japan decides on September 18. Those are two policy tightenings, in the world’s largest and third largest bond markets, inside the same five day window, in a market where the ten year JGB has already crossed 3% for the first time in about three decades and where the yen carry trade is a meaningful part of how leverage in global risk assets is funded. A yen that was at forty year lows in July and required joint intervention is now up sharply on repatriation flows and rate expectations, and Tokyo just demonstrated what that does to an equity index that had spent Monday celebrating an inventory number.

We are not calling for a repeat of August 2024. We are saying that the single most crowded assumption in this market, which is that the AI trade is insulated because its cash flows are contracted, has a funding leg that runs through a currency now moving one percent a day into a central bank meeting. That is a positioning risk, not a thesis risk, and the two get confused most often at exactly this point in a move.

Novartis Has the Worst Week in Its Modern History

Novartis fell about 12% this morning, its steepest single day decline since March 2020, after del-desiran missed its primary endpoint in the Phase III HARBOR study in myotonic dystrophy type 1. That is the third setback in seven days, following pelacarsen’s failure to reduce cardiovascular events in Lp(a)HORIZON despite lowering lipoprotein(a) across 8,323 patients, and a clinical hold on rap-cel after patient deaths. The company reaffirmed 5% to 6% annual sales growth through 2030, which is the correct thing to say and not the thing anyone is trading.

Two failures, three if you count the hold, is not a run of bad luck; it is a readout on how much of the Novartis valuation was pipeline optionality rather than marketed revenue. The Lp(a) result is the more consequential one for the rest of the sector, because it refutes a biomarker rather than a molecule. Amgen was cut to Market Perform on the read-across to olpasiran, and every remaining Lp(a) programme now has to justify why the endpoint should behave differently when the target moved cleanly and the outcome did not. Dyne and Sarepta traded lower on the myotonic dystrophy read. This is a genuinely bad morning for large-cap pharma and a reminder that the sector’s defensive reputation is a function of dividend yield, not of scientific certainty.

Prints, and the One That Actually Matters

Today’s calendar is thin and Thursday carries the week.

GameStop reports after the close, and unusually the numbers are already known because the company pre-announced on August 31.

Our take: the operating business shrank by roughly a fifth on the Switch 2 comparison, planned store closures and the France divestiture, while net income nearly doubled on about $238 million of gains from the eBay derivative and equity position, offset by a $75 million loss on digital assets. A company whose largest disclosed asset is a $4.947 billion stake in another public company is being valued as a holding entity that happens to run stores. Tonight’s release matters only for the operating margin trend inside that shrinking revenue base and for any statement on what the eBay position is for. Everything else was priced a week ago.

Oracle on Thursday is still the print of the week. Consensus sits near $19.13 billion of revenue and non-GAAP EPS guided to $1.72 to $1.76, and neither line will set the stock. Remaining performance obligations will. RPO closed fiscal 2026 at $638 billion, up 363% year on year and up $85 billion sequentially, with management targeting roughly $90 billion of fiscal 2027 revenue against $67.4 billion delivered. The tension is entirely in the funding: fiscal 2026 free cash flow was negative $23.7 billion on $32 billion of operating cash flow because capital expenditure ran to $55.7 billion, and the company expects to raise around $40 billion of debt and equity this year. Roughly $75 billion of that backlog is prepaid or customer-supplied hardware attached to a small number of very large AI contracts.

So Thursday is a referendum on two things at once. Does the backlog keep compounding, which validates the memory, storage and optical bid described above from the demand side; and can it be financed at a bond yield of 4.80% without diluting the equity that is meant to capture it. The stock is down about 18% year to date into a print where options imply an 11% move. That is a market that has already decided the second question is harder than the first.

Geopolitical Risks

Saudi production infrastructure is now the live variable, displacing the strait itself. The specific thing to watch is whether Aramco quantifies the outage; silence past today would tell you more than a number. The Iran and Oman corridor deal is the offsetting event and is described as days away, with an Iranian maritime exclusion zone to be declared alongside it. Both of those resolve in this week, not this quarter.

Canada and the United States have moved from negotiation to reciprocal tariffs at rates up to 50%, with the White House separately telling Bombardier it cannot sell aircraft in the US without domestic manufacturing. There is no visible off-ramp scheduled. Venezuela remains the slow structural counterweight, with the Pentagon holding a 35% stake in NABEP and Chevron working toward 600,000 barrels a day over five years, and none of that arrives in time to matter for the fourth quarter. Germany’s AfD result in Saxony-Anhalt still sits under a European fiscal question with two more state votes this month, and Berlin’s bond market has not yet asked about it. China and Taiwan are unchanged, though the rare earth shipment refusals and the drafted FCC ban on Chinese optical transceivers remain unresolved ahead of Xi’s Washington visit.

Accountability

We owe a correction before anything else. Monday’s note and Friday’s wrap both put Oracle’s remaining performance obligations at $455 billion, up 359%, and described half a trillion as the level to clear. That figure was a full year stale. Oracle reported RPO of $638 billion at the fiscal fourth quarter on June 10, up 363%, having already passed $553 billion the quarter before. The half trillion threshold we set was cleared two quarters ago. The argument survives, because what matters is the rate of change and the funding, but the number was wrong and it was the number the whole case rested on. Corrected above.

Two calls landed. We wrote yesterday that pricing this conflict as a slow physical drain was rational only while the ladder stayed calibrated, and dangerously complacent the moment a strike hit a facility rather than a hull. That happened this morning, and crude took out $99. We also flagged that the memory move had gone from contrarian to consensus fast enough that the next leg was more likely to be given back and retaken than delivered in a straight line. The Nikkei handed back four fifths of Monday within a session. We will not claim full credit for that one, because it came back through the yen rather than through the crowding channel we named, and getting the right outcome through the wrong mechanism is worth recording as such.

Everything above is a pre-open mark. The 9:30 open is the first American price on four days of news, and a tape that has to absorb an oil shock, a tariff schedule and a currency move in the same session should be expected to be wide before it is orderly.