The Daily Spark

Tanker crossings through the Strait of Hormuz have declined again, and refining crack spreads are widening sharply as constrained crude oil supplies prevent refineries from boosting throughput to meet steady demand for gasoline and diesel.

The real tail risk emerges if inventories at critical nodes like airports or power plants suddenly run dry, triggering non-linear cascading damage.

Edward Fishman's Chokepoints thesis explains the dynamic: geopolitical leverage concentrates not in reserves, but in the narrow passages through which the world's oil actually flows.

Our chart book, available here, does a deeper dive into energy demand and supply dynamics around the Strait of Hormuz.

Three events converge next week that could dramatically reshape market narratives:

1) Hyperscaler earnings (July 29-30)

Microsoft, Meta and Amazon report back-to-back, and investors are watching the AI trade obsessively. Are capex investments, earnings growth and returns accelerating or stalling?

2) Fed meeting (July 28-29)

Nine of 18 Fed officials project a rate hike this year. Forward guidance has been eliminated, making individual Fed speakers the only remaining signal of intent. Recent Fedspeak has been unambiguously hawkish on inflation. With no roadmap, markets will parse every word for clues about the path ahead.

3) Iran conflict re-escalating

The Strait of Hormuz is a critical chokepoint. Escalation could constrict oil flows, deplete global reserves, and spike energy prices, adding fuel to the inflation battle the Fed is waging and putting further upward pressure on yields.

Bottom line: Next week answers three separate questions: Are AI capex investments panning out? Can the Fed contain inflation? Will geopolitics stay manageable? Markets are currently betting yes on all three.

Since the Fed raised interest rates in 2022, annuity sales have basically doubled and stayed at that higher level. In turn, these strong sales are creating strong institutional demand for both public and private credit, see chart below.

Six years after the pandemic, weekend subway ridership in New York has recovered to nearly 90% of its 2019 level while weekdays remain stuck near 75%, consistent with Nick Bloom's finding that the shift to hybrid work has permanently thinned the weekday commute, see chart below and here.

The effective tariff rate has declined from 11% at the peak to between 6% and 7% today, see chart below.

With the Fed estimating the non-accelerating inflation rate of unemployment (NAIRU) at below 4.5%, and unemployment having stayed at or below that level for 57 months, tied for the longest such streak on record, the labor market has been operating in excess-demand territory for an unusually long time. That persistent tightness is a key reason inflation has remained elevated: when unemployment runs below NAIRU, wages and prices face sustained upward pressure.

The chart below puts this streak in historical context. Prior episodes of sub-4.5% unemployment were typically far shorter. The current one is one of the longest on record, which helps explain why the ongoing inflation overshoot since 2021 has been so stubborn.

The bottom line is that a strong economy is the reason why inflation has been high, and only by keeping rates higher for longer can the Fed cool inflation down towards the FOMC’s 2% inflation target.

The Epoch Capabilities Index combines scores from many different AI benchmarks into a single "general capability" scale, and the chart below shows that open-weight models trail the closed-weight frontier by around four months. For more, see also here.

The cover ratio measures how many dollars of investor orders a bond deal receives for every dollar of bonds issued. For hyperscalers, it has fallen from nearly 5x in February 2026 to below 2x in July, suggesting investors may need wider spreads to absorb additional hyperscaler supply, see chart below. For more discussion, see also here.

To gauge how much AI affects a job, researchers rate it from 0 to 1 based on how many of its tasks could be, or already are, done with AI. The closer to 1, the more exposed the job.

The trouble is that the studies doing this quantification agree for low-exposure jobs like hairdressers and dancers. But for the high-exposure jobs everyone actually worries about, like tax preparers, telemarketers and mathematicians, they disagree wildly, see chart below.

So the jobs most likely to be called "at risk" are the ones we understand least. These roles tend to involve many different tasks, and while AI can do some, others are hard to automate, which is exactly why the measures disagree.

The Yale Budget Lab came to a similar conclusion here, and my colleague Sania Edlich and I will keep digging into this in upcoming Sparks. For more, see also here.

Net foreign inflows into US equities have surged to a record high, driven in large part by overseas investors seeking AI exposure they cannot get in their home markets, see chart below.

With most foreign equity investors not hedging their FX risk, the bottom line is that if AI disappoints, the resulting pullback in these inflows would be a significant downside risk to the US dollar.