One session, two major sell-offs: the AI trade rolled back over (Oracle's worst week since the 2001 dot-com bust, on AI-FINANCING fears) and the Iran ceasefire broke (the US struck Iran, a tanker was hit in Hormuz). Yet oil fell ~2% - the supply glut still outweighing the geopolitical tail.

Two reversals the book had flagged as risks arrived together on Friday, in what one wrap called 'one session, two major sell-offs.' The first was the artificial-intelligence trade: it rolled back over, and the leadership of the decline changed - Oracle posted its worst week since the 2001 dot-com bust as investors fixated not on chip demand but on the FINANCING of the AI build-out, a deeper and more structural worry than the memory-cycle fear that started the month. The second was Iran: the ceasefire broke. The United States struck Iran after accusing Tehran of violating the truce, a tanker was attacked in the Strait of Hormuz, and a UN agency paused its evacuation plan. The two pillars that had defined the week - a recovering AI trade and a de-escalating Middle East - gave way in the same session.

This is the materialisation of exactly the risks Thursday's brief ranked: the AI recovery proving narrow and rolling over (#3), and the Iran framework slipping (#5). What did NOT happen is the obvious consequence: oil fell about 2% even as the US struck Iran, because the market looked past the fresh tensions and focused on the supply glut the deal has already unleashed - OPEC barrels and returning Iranian crude. That is the non-obvious read of the day: the geopolitical tail re-armed, but the physical supply wave is, for now, the stronger force on the oil price. The risk premium that should accompany a US strike is simply not in the crude.

The shift in the AI narrative is the more consequential one for portfolios. The month's first scare was about whether memory demand was peaking; this one is about whether the hundreds of billions in AI infrastructure spending can be financed at a higher cost of capital - Oracle's debt-funded build is the test case, and its worst week in a generation is the market starting to price that question. A demand worry is cyclical; a financing worry is structural, and it reaches the whole capex chain, not one stock.

The case against reading too much into a single Friday deserves a hearing. Friday sessions into a weekend with live geopolitical risk exaggerate de-risking; the Iran strike may be a one-off enforcement action rather than a return to war; and a single bad week for one debt-heavy name (Oracle) is not yet a verdict on the whole AI-financing model. The de-risking read is wrong if Monday opens calm, if the Iran situation de-escalates again quickly, or if AI leadership stabilises on the still-intact demand picture.

Net, conviction is high that the week ended with both of its supports - the AI recovery and the Iran de-escalation - broken, and that the AI story has shifted from a demand question to a harder financing one. Conviction is medium on direction from here, because the single most important price - oil - is refusing to confirm the geopolitical danger, and Monday's reopen into an unresolved strike is the real test. The read to carry into the week: watch whether oil finally prices the Iran risk, and whether the AI-financing fear spreads beyond Oracle. NOTE: Friday's market levels are from reporting; the 26 Jun session is not yet in our market-data feed (last loaded: 25 Jun).