The War Is Over. The Floor Isn't.

The war is over, or close enough that the market is trading it that way. The United States and Iran have signed, the first tankers have cleared the Strait of Hormuz, and Brent has fallen hard, into the high 70s from the mid 80s in a matter of sessions. This is the de-escalation the whole curve was waiting for. And the long end of the US bond market has barely moved. The 30-year sits at 4.90%, almost exactly where it traded through the worst of the conflict.

The financial press has noticed, and this week it asked the question out loud: the war is over, so why are bond yields not lower? It is the right question. The desk answered it before it was asked. The floor under the long end was never the war premium. It is the bid. And in the space of one week, the bid got thinner while the deficit got heavier.

What the peace did, and what it did not

The relief landed exactly where a war premium should land, and nowhere else. Brent fell close to 5% on the week as the ceasefire took effect and tankers began moving, with the energy complex pricing a partial reopening. That is the war premium leaving the oil price. It is real, and it is welcome. It is also not yet clean, and the desk treats it as conditional: follow-on talks in Switzerland were postponed after Israeli strikes on Lebanon, and the US vice president cancelled his trip. The relief in oil is real; its durability is not yet proven.

It did not land in duration. The 30-year held 4.90%. A genuine war premium in the long bond would have come out with the oil price. Instead the long end sat still while crude dropped, and the front end actually rose: the 2-year climbed to 4.19% after the Federal Reserve, in Kevin Warsh’s first meeting as chair, held rates at 3.75%, stripped the cutting bias out of its statement, and published projections that now point to at least one more hike rather than a cut. The Bank of England told the same story from the other side of the Atlantic: it held at 3.75% on a 7-2 vote, with two members dissenting in favour of a hike. Two major central banks, leaning the same way, into a falling oil price.

So the curve did the opposite of the peace-rally script. Oil fell, the front end rose on a hawkish Fed, and the long end refused to drop. If the long end were about the war, this is not how it would trade.

Why the floor held

Because the level of the long end is set by who has to buy the bonds, and that buyer base is shrinking at the worst possible time.

The US Treasury has to sell a large and rising amount of debt every quarter. For years a deep base of price-insensitive buyers absorbed it: foreign central banks, reserve managers, and above all Japanese institutions recycling the proceeds of near-zero domestic yields into US duration. That base is thinning. The Bank of Japan has raised rates to a 31-year high and is buying fewer of its own bonds, giving Japanese capital a reason to stay home. And this week the pressure sharpened: the yen slid past 161 toward a four-decade low, reviving talk of official intervention. If Japan moves to defend its currency, it does so partly by selling reserves, and the largest foreign holder of US Treasuries turns from a buyer into a seller. The bid does not just stop growing. It can reverse.

Then the supply side got heavier on the very same news. As the war premium left the oil price, the war bill arrived on the fiscal ledger: the Pentagon has reportedly asked for an $80 billion supplemental to cover the conflict’s costs. The deficit that holds the long end elevated did not shrink with the peace. It grew. Peace did not lower the long end. It added to the thing that holds it up.

On the record

This is not hindsight. Each leg was published, dated, before the press asked the question.

Where things stand

Market levels: Treasury yields are the latest daily reads (18 June 2026); FX and gold are live indicative intraday around the morning of 19 June; Brent references the most recent intraday read and is indicative. Not official closes.

The desk’s read

The peace dividend is real, and it is landing where it should: in the oil price, in gasoline, in the energy line of future inflation. It is not landing in the long end, and it was never going to, because the long end is a bid problem and a ceasefire does not buy a single Treasury. This week made the point twice over. The thing that has historically absorbed US supply, the Japanese bid, is not just thinning but edging toward the level where it could reverse. And the deficit that underwrites the floor just absorbed a fresh war bill. Two central banks leaned hawkish into the same tape. Through all of it the 30-year sat on its floor.

One read worth stating plainly: gold. With the war premium gone, gold has stopped trading the conflict and started trading the dollar, and the same dollar bid that is the relative winner of this regime has pulled it off its highs. Our earlier framing leaned on gold as a fiscal hedge; in the event the dollar has been the stronger force. From here the read is gold capped and biased lower while the dollar bid holds. It does not change the rates picture, which is the heart of the note.

What to watch

First, the next Treasury refunding and the indirect-bidder share at the coming long-end auctions. That is the single cleanest test of whether the foreign bid is holding or retreating, and it is where the whole thesis is either confirmed or broken. Second, the yen: a move to defend it is the mechanism by which the largest foreign holder becomes a seller, so watch USD/JPY and any intervention. Third, core PCE on 25 June, the Fed’s preferred gauge and the next read on whether the hawkish dot is heading from conditional toward probable. Fourth, the 30-year itself: a break above 5.15 says the buyer strike is arriving; a break below 4.85 says the bid is coming back. Fifth, the durability of the ceasefire itself: the postponed Switzerland talks and the Israeli strikes on Lebanon are the reminder that the energy relief rests on a framework that is signed but not yet secure. Until then, the floor is the regime.

The war is over. The floor is not. We read the data. We call the paths.