I write about this war once a month, mostly for myself. Writing is how I think. Building an argument on the page is how I work out what might be happening and what might come next. The method earns its keep, but it keeps catching me out, because this war has surprised me more than once.
The cleanest miss was the last one. In June I argued the peace would hold, albeit messily, on two self-enforcing ceilings neither side could afford to break. It broke inside three weeks. I had the deal itself roughly right, having put a negotiated climbdown in May at better than three chances in four, and the memorandum arrived on cue on 17 June. What surprised me was how briefly it lasted, and that Iran discarded a settlement written largely in its favour.
Put the misses together and a pattern shows through. I keep expecting this thing to end. Brevity in March, a deal by summer in May, a durable settlement in June. The one outcome I kept discounting, the long grind, is the one we now have. It is becoming a morass.
There is a second habit sitting underneath that one, and I only caught it while writing this month's piece. I keep assuming the staff work has been done. In March I assumed serious planning lay upstream of the military execution. It did not. This month I assumed an agreement had been tested for workability before it was put to Iran. It had not. Both are the same error in different clothes. I keep crediting the US in this war with more competence than it possesses.
So I will pay less attention this month to what Tehran and Washington say, and more to what they do. This piece is mostly about the memorandum of 17 June, because a peace that died in eight days is the most instructive thing that has happened in this war. Understand why it could not survive and you understand why the morass is stable, and why the next attempt is likely to fail in much the same way. I close with the energy market, which is splitting in two, and with the forces that will shape the months either side of the November midterms.
The state of play
The June memorandum is finished in all but name. Iran chipped at it within a week of signing. Washington pulled Iran's oil licence on 7 July. Trump called the deal over on the 8th and 9th, at a NATO summit of all places. On 11 July the Revolutionary Guard formally declared the Strait of Hormuz closed and fired warning shots at a ship it said was using an unapproved route.
And yet. On 25 July, after nearly two weeks of heavy strikes, Trump quietly halted the bombing, and Oman sent a delegation to Tehran the same night. Talks accelerated over the days that followed. So the honest description is not that war has resumed or that peace has collapsed. It is oscillation. Strike, pause, talk, repeat. The speed of the collapse surprised me; the equilibrium did not. It is more fragile than I thought a month ago, and it could still break either way, collapsing into something worse or tightening into something more disciplined.
The deal died on arrival
The puzzle has nagged at me since June. Why sign a deal only to abandon it within days? Signing bought Iran real things. Sanctions relief, an oil licence, the end of the blockade, unfrozen money, a reopened Strait on paper. It threw most of that away inside a week for the sake of a few drone strikes.
I have been explaining this with internal Iranian fragmentation, and I no longer think that is the answer. It is close to unfalsifiable, and it is what we always reach for when a state does something we cannot parse. There is a better explanation sitting in the price data, and it is a good deal more rational. Iran did not walk away because it was incoherent. It walked away because the bargain it had provisionally accepted proved even less workable than it had feared, and it took only days to establish that.
The objective was never in doubt. The memorandum promised free passage for 60 days only, and that little word left the door open to permanent tolls worth as much as \$40 billion a year, a revenue stream Washington cannot switch off the way it switches off sanctions relief. My reading is that Iran signed to bank the concessions and to establish, in writing, that the Strait is a thing you pay to cross.
Iran held one asset, the ability to shut the Strait, and it had put Brent above \$126 on 30 April. It agreed to stop using the asset for sixty days in return for the legal right to sell its oil. A chokehold exchanged for a cash flow. On paper that is a reasonable swap.
But Iran was not failing to sell oil for want of permission. It had been selling oil illegally for years, through an entire parallel apparatus of shadow tankers, deep discounts, non-dollar settlement and Chinese independent refiners. To collect on the waiver it had to migrate out of that grey system and into the legal one. That migration is the mechanism, and it is where the whole thing came apart.
The legal channel existed on paper but never became commercially usable. The memorandum was signed on 17 June. The instrument meant to implement it, OFAC's General License X, did not appear until 21 June. The commercial promise came first and the legal machinery was reverse engineered afterwards. That is the wrong order, and it tells you the terms were never tested against what Washington could deliver.
When the licence did arrive it was close to unusable. It carried no carve-out for transactions touching the Revolutionary Guard, which is commercially fatal. The Guard is embedded in Iranian ports, shipping and the oil trade, and retains terrorism designations in several jurisdictions, so a bank running ordinary diligence is very likely to strike an IRGC connection and decline. The further State Department waivers that specialists judged necessary never appeared. OFAC published no compliance guidance. And part of the problem lay beyond Washington's gift altogether, because European and British sanctions remained in force, which keeps Iran off SWIFT whatever an American licence says. Sanctions specialists judged the licence difficult to use, and said industry was right to be wary of relief granted that quickly. Within days of its issue the Atlantic Council was calling on Treasury to add compliance guidance, comfort letters and fact sheets to coax the banks off the sidelines. The scaffolding had not been built.
The licence did authorise payment in dollars, so the Treasury Secretary's announcement was accurate and, in practice, close to worthless. What Iran had been promised was not permission to sell, which it had never really lacked. It was promotion into the legitimate financial system, and that never came. Trade could still creep through the old workarounds, yuan routed via a Chinese branch of an Indian bank, rupee and barter arrangements of the sort used after 2019. Those cost more, and they were available to Iran anyway. So the deal handed Iran nothing on payments it did not already possess, while taking from it the one asset it did.
And the grey channel was worth less than it had been. The Chinese independents bought Iranian crude because it was cheap, and it was cheap because it was illegal. Legalisation narrows that discount by moving the price toward market rates, which removes much of the reason to bother. China was buying less in any case, for reasons unconnected to the memorandum, having stockpiled through 2025 and cut imports generally once the Strait closed. Its share of Iranian barrels had slipped to just under 11 percent by May, from about 14 across 2025. Iran was being invited to step off one platform onto another that had not arrived, and the platform it was leaving had itself been lowered.
The export figures are the thing most easily misread. Cargoes did leave, and one estimate puts the value near \$5 billion. But an export figure is barrels multiplied by price. It is booked, not banked. It tells you hulls departed. It says nothing about what cleared, when, at what discount, or through how many intermediaries. Iran could watch its own crude sailing away, read a healthy number in the press, and still be looking at receivables running through the same expensive plumbing it had been using for years.
The money was the other half of the bargain. Point 11 of the memorandum has the United States undertaking to make Iran's frozen funds fully available for use. The same clause then provides that the two sides will mutually agree on the procedures for release during the negotiation. That is an undertaking to reach an undertaking, and it collapsed on contact.
On 23 June, the first day of talks, Iran announced that \$12 billion had been agreed. Washington told a different story. Trump's version was that the money would go on American farm produce, that the food would be bought exclusively through the United States from our farmers. Iran's foreign ministry answered that the assets would be employed with absolute liberty by Iran to buy whatever it needed. The next day the Treasury Secretary confirmed that his department would oversee the funds whenever they were finally released. Iran had signed for unfrozen money and was being offered a supervised grocery account.
The same day produced a second discovery. Vance announced that Iran had agreed to admit IAEA inspectors and called it a major milestone, and Trump posted that Iran had fully and completely agreed to the highest level of nuclear inspections. Iran's foreign ministry said there was no clear schedule for any such thing. Six days after signature, the parties could not agree on what they had signed.
Now assemble that week. The licence landed on 21 June and could not clear a bank. On 23 June the frozen funds turned out to be an agreement to agree, and a supervised one at that. On 24 June the Treasury confirmed it would hold the purse. On 25 and 26 June Iran struck ships in the Strait. Both parts of Iran's compensation failed in public within six days of signing, and Iran defected within forty-eight hours of the second failure.
The contrary case was made at the time, and made well. Two weeks in, Keith Johnson judged in Foreign Policy that the memorandum was paying off early for Iran: relief through the summer, exports rising, written commitments on the frozen assets, and Washington's own leverage draining away. Every element is accurate. But those are terms obtained rather than value received. The rising exports followed the lifting of the blockade, not the granting of the licence, and we have just seen what the commitment on frozen assets was worth six days later. Nor need Iran have received nothing, because a bargain can pay something today and still be worth abandoning if every further day surrenders more leverage than it delivers value. If Iran was winning, its move was to sit still and collect. It did not.
And while all this was going on, the meter was running hard the other way. Brent fell from its April peak to \$70.82 by 2 July, down more than 38 percent and back to pre-war levels, with daily crossings of the Strait recovering from 22 to 40. The asset Iran had pledged not to use was being repriced toward nothing precisely because it had been pledged. Compliance was self-liquidating. The better Iran behaved, the less its concession was worth, and the weaker its hand would be when the sixty days expired and the real negotiation began.
The deal converted Iran's leverage into a cash flow, except that the leverage converted instantly and the cash flow never arrived.
Which makes the timing intelligible, and tighter than I first supposed. I had assumed Iran needed a week or two to find that payments were not clearing. It needed no such time. The licence was published on 21 June and Iran struck ships on 25 and 26 June. Any competent counterparty could read that document and see that the promised channel was unlikely to work. Iran did not need to learn it through experience. The conclusion was available immediately from the text.
Nor would patience have rescued anything. At day fifty-nine Iran would have reached the table with sharply diminished leverage, little dependable new revenue, and a waterway the world had relearned to use without asking its permission. Iran did not renege on a good deal. It escaped a bad one, and it was slightly late doing so.
When Iran struck a cargo ship near Oman on 26 June, crude kept falling anyway. Tehran had just discovered that its principal weapon no longer moved the crude market price. A fortnight later it hit three vessels rather than one. That is not randomness. That is an actor increasing the dose because the first dose failed to register, and it predicts continued escalation until something finally does.
None of this requires bad faith. Bad faith would mean Iran knew at signing that it would breach. But a party planning to defect does not negotiate an explicit clause preserving free passage for 60 days only, which is a claim written into the text for the period afterwards. You telegraph a right you mean to assert. You do not telegraph a breach you mean to commit. Nor does a state looting a window export at half its capacity.
Which suggests that the sincerity of the parties was never the point. The useful question is whether the agreement was self-enforcing, meaning that compliance beat defection for both sides at every moment of its life. This one had ceased to be self-enforcing for Iran before its first week was out. Nor was it self-enforcing for Washington, which handed over a revocable licence and duly revoked it on 7 July, and which was watching its own leverage drain as the ceasefire ran, its sanctions spent, its blockade lifted and its munitions depleted. Both sides were losing ground by standing still, which is why the first plausible pretext was seized by each of them.
My own view is that neither side was pure and both wanted the thing to hold. Washington pushed hard, pressing every advantage, attaching conditions and keeping the licence revocable. It did not want the failure it got. But good faith is not only a matter of intent. It requires process. You cannot promise in good faith what you have not established you can deliver, which means the vetting failure is not a separate charge from the credibility one. It is the same charge. Signing a document nobody has checked for executability is a way of making promises you do not know you can keep, and the party across the table experiences that as bad faith whether or not it was meant that way.
Iran had reason to test early. The United States walked away from the nuclear agreement in 2018 and reimposed the sanctions it had lifted, so Tehran already knew that American relief could be taken back. This war added to the ledger. I wrote in May that Trump's public statements about it had a hit rate that was low and declining, and that they were signals about him rather than about the situation. Supporters forget. Foreign ministries file. By 17 June the currency he was offering Tehran was one he had spent years devaluing.
That changes the sequence I have been describing. I have been writing as though Iran signed, then discovered the promises were hollow, then left. More likely it signed already doubting them, because it needed the relief and had nothing better on offer, and then watched closely for the first sign that its doubts were sound. Confirmation took six days. A licence its banks could not touch on the 21st, a supervised grocery account on the 23rd, and Washington announcing inspection commitments Tehran denied making on the same day. Iran was not learning. It was checking.
And a party carrying that kind of deficit does not get to deal in promises. The burden sat with Washington to perform immediately, visibly, and in a form Iran could verify for itself, precisely because nothing it said would be taken on trust. That was the one route to making this agreement hold, and it required exactly the front-loaded, irreversible delivery that nobody had checked was possible. Washington managed none of it.
Which is why the alacrity of the rupture is not a puzzle but a piece of evidence. A party with no priors takes weeks to conclude that a counterparty will not perform. A party that already expects non-performance needs only the first instance to act on it. Eight days stops looking fast.
The bill arrives with the next attempt rather than this one. Where promises cannot be relied upon, only structure remains.
I still think Tehran is fractured, and the fact that the Revolutionary Guard rather than the foreign ministry declared the Strait closed tells us something real. But fragmentation explains the untidiness, not the decision. What the decision reflects is a bargain Iran had no business signing. It bargained for permission when its problem was payment. And it made the whole arrangement depend on the willing cooperation of institutions that were never in the room and had no reason to help, namely the compliance departments of American banks. No signature could deliver those.
Washington's part is not much better, and it failed in the opposite direction. An agreement of this kind is normally run past every agency that will have to implement it, and past Treasury above all, before it is put to the other side. The evidence suggests that did not happen. The licence followed the signature instead of shaping it, and when it came it was unusable in ways any sanctions lawyer would have anticipated. The pattern repeats across the document. The passage clause bound Iran to its best efforts, with no charge for sixty days only. Point 11 promised the funds while deferring the procedures for releasing them. These are agreements to agree, which is what a negotiation produces when nobody has asked how each component will be executed. It has the look of envoy diplomacy rather than an interagency process. And it means the defect was not merely that implementation went badly. The memorandum had postponed the implementing terms, so implementation was the bargain.
I should be careful how hard I lean on that, because I am reasoning from the instrument rather than from an account of the room, and no reporting I have seen states that Treasury was cut out. But the artefact is eloquent. You do not promise a counterparty that it may sell its oil and then discover, four days later, that your own licence cannot clear a bank.
I doubt Trump understood what he was holding, and I want to be fair about where that fault sits. No head of government reads the implementing detail, and none should have to. The job of the machine beneath him is to confirm that the agreement has been vetted and that every component can be delivered. That assurance is most of what a government is for. Either nobody offered it, or somebody offered it without grounds. The waiver was revocable and time-limited because that is how Treasury writes every waiver, not because anybody designed a snare. So Trump did not build a trap and mistake it for a peace. He built a peace, it functioned as a trap, and he appears to have been as surprised as anyone when it sprang. Washington then compounded the injury on 7 July, replacing the licence with a wind-down of ten days for transactions whose participants had been promised sixty.
From Tehran this would have looked like bait and switch, and that lens probably comes closest to explaining the walk. The pattern is a familiar one from Trump's commercial career. Close the room. Get the signature, announce the deal, push the hard terms into a later negotiation. In property and entertainment that is a technique rather than a fraud, because contracts are enforceable and both parties expect to trade again next year. Transplanted into statecraft it becomes something else. Deferring the operative terms moves the real bargaining into a window during which one side's leverage is collapsing. Iran's concession was immediate, observable and self-devaluing. America's were deferred, supervised and revocable. So an undertaking to settle it later was never neutral. It meant settling it later, when you are weaker. Washington pushed the deferred terms as far as the language would bear. The farm-produce condition and the Treasury supervision were attempts to make Point 11 mean considerably less than Tehran thought it meant. That was strategy, not fraud. But the structure is what Iran was staring at by 23 June.
Which is why I would hesitate to call this a negotiation at all. States strike unenforceable bargains all the time, so that by itself is not the complaint. The complaint is the asymmetry, and Iran was made to pay for the announcement.
Washington was happy enough to walk as well. When Iran breached, it made no visible attempt to save the agreement, though the administrative repair had already been set out in public, being the compliance guidance, comfort letters and fact sheets the Atlantic Council had called for within days of the licence appearing. None of it arrived. Washington revoked the licence instead and pronounced the memorandum dead within forty-eight hours. That was not insincerity. Washington needed this agreement, having no military route to opening the Strait and no appetite for ground troops against something like two-thirds public opposition. It is easier read as a scramble to save face, because revoking the licence made the rupture Iran's fault, which was a good deal more comfortable than conceding that the thing had never been capable of working, and handing fresh concessions to a state that had just struck commercial shipping was impossible three months out from the midterms.
Both sides had in truth discovered the limits of their own coercion, and neither could say so. Iran hit a ship on 26 June and the oil price fell anyway. America hit ninety targets on 8 and 9 July and Iran answered by declaring the Strait closed on the 11th. Walking away was easier for both than admitting that, which explains the lack of any rescue attempt better than any theory about who meant what in June. No villainy is needed to explain the wreckage. But the practical consequence outlasts the diagnosis. Unless that asymmetry is inverted, the next attempt fails in the same way. Watch whether Iran's benefits arrive front-loaded and irreversible, meaning assets actually transferred rather than permissions granted. If they do not, the sincerity of everyone involved will make no difference at all.
Reduced to its bones, the United States negotiated an agreement it could not implement. At signing it did not know that, because it had not done the work any competent government would have done. Iran walked away when it saw that it had surrendered its leverage and was unlikely to be paid what it had been promised.
That is what makes the morass durable. Neither side can restore the old bargain, and neither has the coercive power to impose a new one.
Crude cools, gas bites
The energy market is telling a story in two halves that most of the coverage runs together.
Crude has come off the boil. Oil remains elevated on the short peace, but it is not as high as it was earlier this year, and it keeps easing because the world is learning how to reroute it. Tankers change course, other producers lift output, and the barrels find their way to market. On oil, then, the world is showing it can live around Hormuz, and that half of Iran's leverage decays every month the war drags on.
Gas is a different animal, because you cannot reroute a cargo of liquefied natural gas the way you reroute a barrel of oil. It needs plants to chill it and terminals to receive it, and you do not build those in a season. Qatar is the world's second-largest exporter and about a fifth of global supply, and it ships through Hormuz. Its damaged capacity is under force majeure into October, with no full recovery expected before the back half of 2027. The numbers are hard. European storage sits near 50% in late July, against a target of 80% by 1 November and a five-year norm closer to 90%. The benchmark price has jumped past €60, up more than half since the June deal. Even if Qatar came fully back by the end of September, the best guess has European storage reaching only about 75% by November. And Asia is outbidding Europe for every spare cargo, so Europe cannot simply spend its way clear of a physical shortage.
A price is not a forecast. It is a distribution, weighted toward the likely and padded for the catastrophic. So €60 does not say Europe will freeze. It says the market is paying to insure against the chance that it does, and that the insurance has become expensive. Read the crude side the same way. Oil easing is not traders calling the war over. It is the closure premium deflating as the market watches rerouting work. Which makes the divergence itself the signal worth having. Crude and gas carry the identical geopolitical risk, so when one premium deflates while the other inflates, the market is not contradicting itself. It is telling us which chokepoint has a substitute and which does not. Iran has been running the same experiment, at rather greater cost, and it now has its answer.
A long cold European winter is the wildcard, and it favours Iran. One consultancy warns that a closure lasting into the autumn could tip Britain and Europe toward recession by year end. Europe's escape in 2022 was a run of mild weather, and mild weather is not a plan. So the thesis I have carried since March, that Iran's prize is a melting asset, needs cutting in two. The oil toll booth sits on a road people are leaving. The gas chokehold sits on a road with no detour before 2027. Iran's real leverage is not crude, and it is not even the toll. It is the power to keep European gas scarce through one northern winter, and it is aimed less at the American petrol pump than at the seams of the Western alliance.
Which force prevails
That leaves a set of forces, and the outcome turns on which prevails. The electoral calendar disciplines both sides until November. Trump wants petrol cheap and the war quiet before the vote, and his congressional authorisation ran only sixty days, so a longer campaign means asking an exposed Republican majority for more. Iran reads the same calendar and leans on its proxies to raise American casualties and pump prices while his hands are tied. I expect that discipline to fray as winter nears and the gas leverage climbs.
Winter runs for Iran. Its grip on gas tightens through December and into February. If the cold cracks the European end of the coalition, pressure to reopen the Strait on Iran's toll terms will rise with every hard frost.
There is an assumption buried in that, and it is the weakest link in my own argument. It assumes economic pain converts into political concession. Europe's recent history says otherwise. In 2022 the continent absorbed an energy shock worse than this one and did not fold to Moscow. Governments can choose recession over capitulation, and that year they did. If Europe treats paying Iran to unlock the Strait as a security surrender rather than an energy transaction, then winter buys Tehran a great deal of European suffering and remarkably little leverage.
Trump's free hand is the one I find least predictable. A lame-duck president after November, with his interceptors restocked, is less constrained, not more. Iran's winter leverage peaks in the very same weeks. Two pressures landing on one window is where a miscalculation drags the war wider, and I would not rule it out.
Crude, alone among them, runs against Iran, and it is the reason none of this settles in Tehran's favour by default. Oil says the world can wait Iran out. Gas and a cold winter say it may not have the months to spare. I have stopped predicting when and how this ends. But I will offer one line for the file, to return to later, as is my habit. Do not call Hormuz a dead asset until the northern spring.
No comments:
Post a Comment