Corporate America has delivered seven consecutive quarters of double-digit profit growth, and consensus is looking for roughly 25% year-on-year growth in Q3 S&P 500 earnings. When the earnings engine is throwing off that kind of torque, investors will tolerate a surprising amount of macro gravel in the gearbox. $100 oil becomes a nuisance, 5%+ yields become survivable, and AI capex gets treated less like borrowing and more like buying future monopoly rents.

Takeaways by Dark Side of the Boom™

• Stocks are still near the highs because earnings remain strong enough to absorb $100 oil and 5% yields, but that cushion is getting thinner.

• Oil, Treasuries and AI are converging on the same problem: an increasingly expensive future needs increasingly expensive capital.

• Credit is starting to discriminate before equities do, with junk spreads and AI-linked CDS flashing more stress than the index.

• Next week’s earnings season becomes the real test of whether profits can still outrun the rising cost of money.

Hormuz, Hurricanes and Hikes

Wall Street spent the week walking through a burning kitchen carrying a tray of champagne, and somehow still made it to Friday without spilling much.

The S&P 500 added 0.6% Friday, the Dow 0.8% and the Nasdaq another 0.6%, leaving the major averages within touching distance of their records even with Brent settled at $104.72, the 10-year Treasury around 5.24%, shipping risk still alive around Hormuz and the dollar heading for its longest weekly advance since early 2025. That is a fairly hostile backdrop for expensive equities, yet the index keeps behaving as though someone else will pick up the tab.

That is the bargain holding the market together: the numerator has been so strong that investors have been willing to ignore an increasingly expensive denominator.

This week, however, the denominator started asking for attention.

Oil went first. Brent spent five days trading missiles, tankers, hurricanes, election timing and diesel diplomacy as though somebody had wired the futures curve directly into the geopolitical newsroom. Hormuz remained messy, vessels were still being hit, Hurricane Isaias shut in a large chunk of Gulf production, and then President Trump tried to lower the temperature by promising no attack on Iran before the midterms while also unveiling a deal with Russia to bring more diesel into the market.

The immediate war premium came off, but the barrel itself refused to behave as if peace had broken out.

Washington can turn down the thermostat with a headline, but it cannot repair the furnace. Tankers still have to sail, refineries still have to run and physical barrels still have to arrive where they are needed. When Brent can settle above $104 even after the strike clock is pushed beyond the election, the market is telling you this is no longer just fear premium. There is still enough physical tightness underneath the story that every fresh disruption lands harder than it should.

Isaias was a good example. In a normal year, shutting in roughly 63% of Gulf crude production would be a short-lived weather trade. In this market, with products already stretched and diesel still doing the heavy lifting in the inflation basket, even a temporary outage hits like somebody kicking the leg out from under a table already carrying too much weight.

The barrel gets the headline, but products still send the invoice.

And the invoice went straight to Treasuries.

The 10-year pushed toward 5.36% during the week and the 30-year traded above 5.70%, yet the auction calendar itself was hardly a train wreck. Buyers showed up. The 10-year sale was strong. The 30-year was serviceable. The Treasury market absorbed the supply and then sat there like a mule in the doorway.

That is the market repricing the rent on money.

Washington needs capital. AI needs capital. Data centres need capital. Power generation, grid upgrades and chip fabs all need capital. Everybody has turned up at the same bank window with a very large funding request, and the market is simply charging more for the privilege.

This is why the real-rate move matters. If most of the selloff is being driven by real yields rather than breakevens, the bond market is not merely shouting “inflation.” It is saying capital itself has become scarcer and more valuable. The Treasury can run buybacks and clear some old duration off the shelf, but that is housekeeping, not magic. You can move furniture around the room; you cannot make the room larger.

The Fed is caught in the middle of that problem and, for once, it is not really the star of the show.

The minutes remain hawkish enough, most officials still think another hike may be required by year-end, and the labour market remains awkward rather than weak. Hiring has slowed, but firing has not really started. That low-hire, low-fire equilibrium is exactly the sort of thing that keeps the Fed from blinking while oil sits north of $100.

But the bigger point is that monetary policy cannot fix the thing now driving the capital cycle.

AI is creating enormous demand for electricity, chips, data centres and financing, while higher rates are squeezing every rate-sensitive industry that does not enjoy a 20% or 30% return on compute. The Fed can lean on demand, but it cannot manufacture more power or GPUs. It can make money more expensive, but it cannot build a substation.

If supply is the bottleneck, higher rates are a brutally expensive way of proving it.

And that is where AI stopped being just a tech story this week and became a macro story.

For months, investors were happy admiring the cathedral. Thursday, somebody finally wandered around the back and asked who financed the scaffolding.

The report that OpenAI’s annualized revenue was closer to $50bn than the $68–70bn figure that had circulated was enough to punch semiconductors, Oracle, Broadcom and the rest of the infrastructure trade in the mouth. The later gross-versus-net clarification and a more reassuring year-end forecast helped tech shake it off, but the scare mattered because it exposed how sensitive the whole complex has become to anything that touches the revenue line.

For months, the market has been willing to tolerate the denominator because the numerator looked bulletproof. Huge capex, high borrowing costs and increasingly elaborate financing structures were all waved through because the expected returns from AI looked extraordinary.

The financing underneath the buildout is getting enormous. Broadcom is looking for tens of billions in private debt, Oracle is exploring off-balance-sheet structures, and AI-related borrowing is beginning to compete directly with Treasury supply for the same pool of savings. The winners can still pay high-single-digit funding costs because the expected returns remain exceptional.

Everybody else is finding out that the landlord has raised the rent.

Credit is where that distinction is starting to show first.

Oracle CDS has pushed to record levels, Broadcom’s has widened sharply and CCC spreads are around multi-year wides, while investment-grade spreads remain calm and the VIX still looks as though somebody slipped it a sleeping tablet. That does not mean a credit accident is around the corner, but it does mean the market is already separating borrowers who can carry today’s cost of capital from those who cannot.

Equities are still drinking champagne upstairs while credit is downstairs checking whether the basement is taking on water.

That is also why the breadth story matters. The S&P can sit near record highs while a huge share of its members are already in correction territory because the winners still have enough muscle to hold the roof up. The danger is not that the roof collapses tomorrow; it is that fewer beams are doing more of the work.

France offered the sovereign version of the same trade. OATs stabilised after the previous week’s stress, but stabilisation is not repair, and EUR/USD sliding toward 1.1160 told you some of the fiscal premium had migrated directly into the currency. Once the bond market stops giving you the benefit of the doubt, the refinancing arithmetic becomes merciless. Brazil, interestingly, showed the mirror image: assets exploded higher on the prospect of greater fiscal discipline, a reminder that credibility still commands a very real premium when investors think it is about to improve.

Then there is the flow number that probably says more about investor psychology than the index does: $166bn into money-market funds in a single week.

That is not capitulation. It is optionality.

Investors are not leaving the casino; they are standing by the cage with chips in hand, waiting for better odds.

That is a very different message from outright fear. It says the market still wants risk, but it is no longer willing to chase every asset at every price. In a world where cash yields something again, patience has become an asset class.

Next week is not simply another round of beats, misses and carefully worded guidance. Corporate America now has to prove that the profit machine remains powerful enough to justify $100 oil, 5%+ long yields, rising AI financing costs and a credit market that is starting to ask harder questions.

The S&P still says the cheque will clear.

And $166bn has gone to sit in the car with the engine running.

For months, the market rewarded ambition. The next phase is about whether the cash flows can actually carry the ambition.

For now, equities still have the benefit of the doubt.

Running Update: Two Weeks to Luang Prabang

I’m officially two weeks out from the Luang Prabang Half Marathon in Laos on October 24, which promises to be one of the most picturesque runs I’ve done in years. But there’s a slight hiccup in the training schedule.

If you’ve been following the flooding here in Thailand, you’ll know the weather hasn’t exactly been cooperating. Throw in a scheduled trip to Hua Hin, and my last three weekend long runs have gone missing in action. Not exactly how you want to arrive at the business end of a training block.

So today is effectively my last proper long run before race day, and at 18 km, it’s going to be more about managing the engine than testing the horsepower.

Last week’s running told me something useful: even at around 80% of my best pace, things start getting uncomfortable beyond 90 minutes. That’s a fairly clear warning not to get carried away when the starting gun goes off in Laos.

Today’s plan is to keep the first 7 km firmly in Zone 2, gradually turn up the dial from there, and see what I’ve got left in the tank over the final stretch. No heroics, no chasing the watch, and definitely no pretending three missed long runs haven’t taken something out of the legs.

At this point, you can’t cram three weekends of endurance training into one Saturday. The fitness is either there or it isn’t. The trick now is finding out how much I’ve got without leaving the race in today’s training session.

And if there’s one thing running in Thailand has taught me, it’s that the heat, humidity and weather always get a vote. Sometimes the smartest race strategy is simply knowing when to back off the throttle.