When the borrower changes, the risk changes

## The week's biggest borrower was not a government

Friday's July employment report was soft enough to end an argument. For a fortnight the question in rates had not been when the Federal Reserve would cut but whether it would raise; by the close, that urgency was gone. Short-dated Treasuries capped their best weekly rally since May, the dollar finished at its weakest since May and the yen jumped a full percent on the print. Equities treated it as oxygen, and the Nasdaq led.

The relief is genuine. What deserves attention is where it landed. Over the same week, Alphabet raised $25bn in a single bond sale — and MarketWatch is explicit about the reason, which is that heavy AI investment is now eating into free cash flow. SoftBank borrowed $10bn against its stake in OpenAI. SK Hynix directed its chip windfall through the Korean corporate bond market. And Nikkei totalled the balance-sheet consequence: the four largest US technology firms now carry $1.46tn of physical assets, a figure that invites comparison with oil majors rather than software companies.

That is a change of funding currency, and it changes who holds the marginal vote. An equity investor prices a multiple and can be patient for years. A credit investor prices a spread, a maturity and a covenant, and is markedly less patient about the distance between money spent now and cash generated later. The evidence that this is already binding is on the buyer's side rather than the borrower's: Japanese issuers have sold more covenant-protected paper this year than in all of last year, a major insurer is contacting companies directly to solicit issuance, and desks are reported to be rethinking AI-linked bond sales altogether. Supply arriving into a buyer base that is asking for protection is how a funding channel gets repriced without a single downgrade.

The second thread of the week is the Gulf, and it rewards the same discipline about what is actually being priced. Our probability layer separates two questions that headlines merge. The live band for a ceasefire by 14 August sits at 0.885; the live band for the blockade ending by 15 August sits at 0.60. Both deadlines are ahead of us, which is what makes them quotable — and we name that explicitly, because on 4 August we cited an expired Hormuz band as though it were current and corrected it in public. Between those two numbers is the whole trade: the fighting stopping and the waterway opening are not the same event. The physical record leans toward the lower figure, with three vessels struck in transit this week and fresh strikes ordered as the war enters a sixth month, even as negotiators close on a draft that would give Tehran oversight of Gulf-bound shipping. And the clearest confirmation of where the premium sits is in the crack rather than the barrel: crude retreated over the week while fuel prices stayed high, because the lost capacity is refining, not crude.

A word on the Federal Reserve, and on our own reasoning. Two stories ran in parallel: the renewed move to remove Governor Cook, alongside unusually direct contact between the President and the chair; and a soft payroll that simply removed the case for a September increase. Yesterday's note argued that the metals were pricing the standing of the institution rather than the level of its rates, and it named a discriminator — the direction of the dollar against hawkish news. Friday did not test that discriminator, because Friday's news was not hawkish. A currency falling on weak data is an ordinary rate move with an ordinary explanation. We are recording that in the open: the structural reading is open, not corroborated, and 12 August is its first clean test. One housekeeping correction in the same spirit — yesterday's published tape carried provisional Friday prints; the settled closes are slightly different, and no direction changed.

What did hold from yesterday is the narrower discriminator inside the metals: silver rose 3.07% while copper fell 1.76% in the same session, a second consecutive day in which the industrial twin declined while the monetary one advanced. Our positioning data sharpens that into this brief's proprietary read, and it points somewhere specific — the speculative futures crowd does not look like the buyer doing the work in silver.