In June I did a rough calculation on what might happen to oil prices once the temporary buffers holding them down disappear. The circumstances in the Strait of Hormuz have changed in recent months, though, which has affected the original estimates. The calculation still indicates higher prices once buffers are exhausted, but the range will be lower than previously estimated if the recent recovery in flows holds, while a fresh disruption to flows would put the market back close to where I had previously estimated it might go.

Several months ago there was virtually nothing transiting Hormuz – the strait was effectively closed, with the subsequent shortfall in global supply mainly covered by SPR releases, lower Chinese imports, bypass pipelines and reduced demand. That’s no longer true. Tanker-tracking data now appears to show direct crude flow through Hormuz at roughly 5-7 mb/d, not the 0-2 mb/d assumed in June. A lot of the confusion around Hormuz numbers has resulted because ‘crude’ and ‘barrels’ measure different things. Pre-war, Hormuz carried about 14.2 mb/d of crude, and around 20.1 mb/d including everything else moving as liquids, such as products, condensate, and similar. So when a big number gets quoted for transit through Hormuz, it’s useful to check which category it’s actually counting. It is also noteworthy that virtually all the improvement has been through Hormuz, not the bypass routes which had been touted as a game-changer but haven’t actually grown much (only ~1.1 mb/d against what they carried before the war).

June’s headline cliff figure – crude and oil products combined, once the temporary buffers were assumed spent – was $136-151 a barrel. Once recalculated on current flows, assuming they hold, the estimated range drops meaningfully to $111-127, yet still a high number and well above the current trading range. If the current elevated Hormuz crude flow drops back toward June’s level, while bypass and current product flows hold, the shortfall returns to roughly what June assumed, taking the estimated price back to around $136-151. That’s not necessarily a worst case scenario – worse combinations are possible – it’s simply what happens if the current flow recovery unwinds.

Oil products haven’t recovered as much as crude has – the shortfall there is still around 3-4 mb/d. June’s $74-95 estimate for prices with buffers still operating excluded the ongoing shortfall in oil products. The updated calculation includes products throughout, alongside revised flows and temporary buffers, producing an estimated current price of $89-114. The difference therefore reflects both the changed physical picture and the consistent treatment of products. The elasticity assumption – how much a price rise cuts demand – is unchanged from before, and matches the underlying research closely enough to keep using it.

US Energy Secretary Chris Wright’s claim of a single-day record of 17 million barrels crossing Hormuz on Monday 31 August isn’t a revised weekly average, and shouldn’t be treated as a new baseline. Kpler, a tanker-tracking firm, has explained why the official and tracked figures keep disagreeing: mostly definitions, not missing or invented oil.

Additionally, LNG has barely recovered, yet my impression is that it is rarely discussed in media or market analysis. Qatar supplied 20% of global LNG supply pre-war, but its exports remain down about 96%. Across the wider Hormuz flow, the estimated LNG shortfall is still 10.5–11 billion cubic feet a day.

The recovery is also more fragile than the daily average suggests. A lot of what counts as “recovered” flow is only visible because it later shows up in a ship-to-ship (StS) transfer – tankers cross the strait dark, then swap cargo onto another ship once clear of it. That crossing work is being carried out by a small, repeatedly-used group of vessels: one shipbroker’s analysis attributed 54% of all shuttle runs since the war began to ships that had already made four or more crossings. Iran is now going after exactly that group. Its blacklist has grown from 45 tankers to 56 in the past week, covering both the ships making the crossing and the larger tankers receiving the transferred cargo, and some major buyers – including several Indian refiners – are already avoiding the listed vessels. The US opened a new front on Tuesday, hitting two Iranian tankers under a declared retaliation policy, after Iran struck two Saudi tankers carrying crude through the strait. Oil moved up on the threat of conflict renewal, with Brent rising about 5% to near $95, its highest since late July.

The price ranges cited above are illustrative calculations, not forecasts. They estimate what the crude price might look like under a set of assumptions. Compared to June the downward direction of price estimates under the same assumptions is apparent, as the current flow recovery has eased the pressure. However, that trend will only last as long as the flow does, and the complex and expensive operation supporting it will continue to be tested, with Iranian escalation testing both the ships doing the crossing and the navy escorting them.

On 20 August, US Treasury Secretary Scott Bessent promised the toughest sanctions in history: the blockade and sanctions would form a one-two punch, and Iran’s regime would break. Four days later he escalated the language further, invoking D-Day itself and promising an economic onslaught that would sever every lifeline sustaining Tehran until it stood entirely alone.

All that Treasury actually delivered on 24 August was some sixty designations covering individuals, companies and vessels, alongside an expansion of the categories of Iranian activity exposed to secondary sanctions and a warning to foreign governments and firms that penalties might follow. No major secondary sanctions were imposed on the institutions that actually help sustain Iran’s economy, and no deadline was set for anyone to stop dealing with it. Reuters accurately reported that Washington threatened penalties but stopped short of imposing them. Daniel Fried, a former US State Department sanctions coordinator now at the Atlantic Council, told Reuters the announcement did not live up to the hype.

The gap between the promise and the delivery is pretty obvious, but why does it exist at all? Because very little is left of Iran’s economy to sanction, and the country sustaining its principal remaining source of export receipts – China – is one that Washington has no intention of confronting.

Iran is already among the most heavily sanctioned economies anywhere. Its principal banks, oil companies, shipping networks, defence institutions and senior officials have been designated for years, in some cases decades. Each new round mostly attaches fresh names to the same prohibited activity – a front company here, a reflagged tanker there. There remain entities to be added to the list, but they are at the margin and, more importantly, very little new economic activity is left to prohibit.

To produce a larger impact would require Washington to move from sanctioning Iran’s disposable intermediaries to imposing costs on the country that actually buys what Iran sells. China took more than 80% of Iran’s shipped oil before the war. Washington has sanctioned a handful of small Chinese “teapot” refiners and intermediaries, but it has stopped well short of sanctioning the major banks that would force Beijing to choose between Iran and access to the dollar system.

Bessent gave away the constraint himself. Asked why the sanctions were not taking effect immediately, he replied: “Why would I want to blow up the global financial system?” He described the delay as a grace period for countries to correct their behaviour, and when asked specifically about Chinese banks, said nobody was beyond reach, but avoided naming one that would actually be targeted. With Trump and Xi due to meet in September, an all-out secondary sanctions campaign against China’s banks was always going to lose out to the wider relationship Washington is trying to manage.

Treasury had already teased something bigger for the days that followed – Bessent told reporters on 24 August to expect a “major announcement” targeting a financial institution by the end of the week. Yet what was announced on 28 August was much smaller. Instead of sanctioning Banque Misr, Egypt’s second-largest bank, the US proposed a rule against just six of its branches in the UAE. Alongside that, Treasury sanctioned one branch manager at Iran’s Bank Melli and a single company in Hong Kong. The UAE branches of Banque Misr allegedly processed $1.8 billion for more than a hundred Iran-linked companies over two and a half years, so the target was not fabricated. However, it was remarkably limited for the promised centrepiece of “Economic D-Day”: it applies only to those six branches, not to Banque Misr itself, and not to any other bank, particularly the major Chinese institutions handling Iran’s most important external trade. Further, it remains only a proposed rule: it must pass through a thirty-day public-comment period before Treasury can finalise it.

That’s not to say the sanctions have no effect. A bank that loses access to a $1.8 billion channel will feel it, and every new name added to the list makes it a little harder and more expensive for Iran to move money and goods. Gradually buyers demand bigger discounts, banks get more cautious, and each transaction takes longer to arrange. That is how sanctions actually work most of the time, through accumulating friction that grinds a country’s trade down over months and years, rather than through one knockout blow. It is a long way, however, from the total isolation Bessent promised, let alone from bringing down the regime.

There is an overarching constraint that I have flagged several times previously which has nothing to do with enforcement. Iran is not a financially open economy that can be starved into a conventional debt crisis: it carries almost no foreign-currency sovereign debt, has no credible exchange-rate peg to defend, imports relatively little compared with most economies, and has no external creditors who can stage a buyers’ strike in rial assets. Its state pays the security apparatus in rials and cannot simply run out of the currency needed to meet that payroll, although inflation steadily erodes what those wages can buy. Iran’s vulnerability is instead physical and distributional: whether Iran can secure enough foreign currency from exports to pay for the imports it still needs to hold off accelerating inflation and shortage. That is not the rapid financial-collapse mechanism that Washington has repeatedly invoked. In the most recent variation, Bessent told Iranian soldiers to recall that “the Berlin Wall fell when ordinary soldiers decided not to shoot” rather than keep enforcing the regime’s orders – an appeal for defection, however unrealistic, not a formula for an economy collapsing.

The naval blockade has probably done more damage to Iran’s economy than any sanctions designation issued since. It attacks the physical vulnerability directly, in a way sanctions on paper cannot. But the blockade also raises energy prices for everyone else, and it cannot force capitulation without being maintained indefinitely or escalated militarily. And both of the latter two options are unappealing for the US from a cost perspective – financially and politically. Thus “Economic D-Day” functions less as a Second Front alongside the blockade than as a way for Washington to present an appearance of escalation while avoiding a further military step and avoiding the admission that the existing approach is not working.

Is that simply a recognition that Washington has run out of options and is hoping the problem simply goes away? It comes close, yet the more accurate description is that Washington is substituting time and rhetoric for a strategy it does not have. Four options remain: strike Iran again, enforce the blockade harder, sanction China’s major banks, or offer Iran enough to reach a negotiated settlement. Each carries a cost the Trump Administration is currently unwilling to pay, yet so does the status quo. “Economic D-Day” lets Washington claim escalation without paying the immediate price of any of those choices while deferring, yet again, the unavoidable decision of how to conclude the conflict on acceptable terms.

Currently the state of affairs in relation to Iran, the US, the Gulf and oil is appropriately reminiscent of that classic military phrase – not FUBAR, although that works too, but instead, “hurry up and wait”. Everything is happening and nothing is happening at the same time (a bit like my day job, which is keeping me from finishing my articles here). The Strait of Hormuz truly is Schrödinger’s Strait.

I am attempting to assess the current state of affairs, trying to cut through and distil all of the information and disinformation, sorting through the opaqueness and uncertainty, and building on my previous analyses. I am nearly finished drafting several pieces which hopefully should provide some useful thoughts on this.

The conflict remains suspended in an uncertain position, with no endgame in sight for any party, and a major reason for that is the goals of both sides appear mutually exclusive. The market appears not to appreciate that and continues to expect an imminent resolution, as it has for months. Crude oil prices should be much higher than they are currently, but have been suppressed (as discussed previously) by temporary buffers. Yet these buffers have a finite life, and the point at which they cease to suppress prices must be approaching.

Despite this, the market, observers and US officials continue to project an overly positive narrative. This reassuring account of the oil market was already prevalent even before the June MOU, but has particularly taken hold from late July. It has focussed on some combination of the usual talking points: global inventories have fallen by less than some of the early forecasts suggested; Saudi Arabia and the UAE have moved more oil through pipelines that bypass the strait; many more tankers are crossing the strait with their transponders switched off; a decline in demand has removed much of the shortfall in barrels; China has cut its imports; production is rising in the Americas; and IEA countries still hold more than a billion barrels in emergency reserves.

A fresh talking point has arisen in the past week: that oil flows through the strait have significantly increased. On 19 August, Axios reported – citing an “unnamed US official” – that a US-run “stealth operation” had been operating for two months, moving nearly 10 mb/d, close to half the pre-war flow. It said 15-20 tankers were entering and leaving the Gulf each night, and that more than 20 million barrels had left during one night alone. The US had supposedly achieved this by creating a protected southern corridor beside Oman, coordinating tankers from a taskforce at Fort Bragg, escorting them with fighter aircraft, and destroying enough Iranian radar and surveillance infrastructure to leave Iran “mostly blind”.

Claims as to the level of flows are a movable feast. US Secretary of Energy Chris Wright initially stated average oil flows were around 9 mb/d, before subsequently giving a figure of 8 mb/d. The claims also seem to have become increasingly unhinged from reality. The latest came on 22 August, when Barak Ravid of Axios reported – citing three US officials – that around 40 tankers had transited in and out of the strait’s southern deep channel on Friday night (21 August), moving roughly 16 million barrels out of the strait in that single night.

Pre-war data indicates approximately 70 tankers transited (both ways) per day carrying ~20 million barrels of crude oil plus liquids. Axios is suggesting that on Friday 40 tankers transited (both ways) carrying 16 million barrels of oil (assuming crude and liquids as well for comparison) and in only the southern route. That means ~80% of pre-war oil flow is transiting through a single, constrained channel compared to the completely open pre-war channel. And through that same constrained channel the implied proportion of outbound tankers that are VLCCs has roughly tripled by my calculations, compared with pre-war, and yet this high concentration of enormous vessels supposedly remains undetected.

Turning off a transponder makes a ship invisible to AIS trackers, but it does nothing about physical visibility. The Strait of Hormuz is one of the most surveilled bodies of water on Earth right now. Just three VLCCs travelling in convoy is more than 1 km in length, not to mention the smaller tankers and USN escort vessels. Additionally, at its narrowest point the strait is only 21 nautical miles wide, so can be physically observed from the Iranian shore, Iranian boats at a distance or Iranian drones.

Further, if this level of flow had been achieved for the past two months, then why is the US still drawing on the SPR, and why has there been no corresponding increase in Gulf loadings, Asian crude arrivals, refinery purchasing or refinery throughput?

To be clear, I am not arguing that the Strait of Hormuz is fully closed – it is not and more oil is moving through it than in previous months. Some tankers are travelling without broadcasting AIS signals. A shuttle arrangement has developed in which vessels collect oil inside the Gulf, cross the strait and transfer it to other tankers waiting off Oman or Fujairah. The US is helping to coordinate and protect at least some of these movements. None of that is seriously disputed.

The issue is how much oil this operation is actually moving, and that is where evidence gives way to increasingly extravagant claims. These are extraordinary claims. They would materially alter the oil balance if true. But every extraordinary part of the Axios story ultimately comes back to the same interested source: unidentified US officials.

The next article will examine these discrepancies in detail, including the vessel calculations underlying the claimed volumes, what commercial tracking data actually shows, and the other physical and commercial traces that movements on this scale should have produced. Subsequent pieces will address the broader reassuring narrative – dark transit, China, rising production and emergency reserves – before turning to the significance of extremely high refining margins and the possible endgames for Iran, the US and the region.

US commercial inventories reported a massive crude build last week. This may be touted as evidence of a rebuild in supply and subsequent oil glut, yet once you look at the underlying numbers there is limited evidence of this. The build consisted of SPR release (35%), identifiable market surplus (36%), transfers of other hydrocarbon liquids into crude supply (9%), and the EIA’s balancing adjustment (20%). The market surplus was fully driven by unusually high imports and low exports.

Based on latest EIA data, US refineries are already running hard, processing 17.2 mb/d, with utilisation at 96.2%. Crude imports jumped by 1.14 mb/d to 7.3 mb/d, particularly because of increased arrivals from Canada and Venezuela, plus delayed Middle Eastern cargoes also began arriving. Meanwhile exports fell by 0.627 mb/d to only 3.06 mb/d, their lowest level since November 2025. Domestic production remained relatively steady at 13.8 mb/d.

So the calculation for the week is:

On a daily basis, the calculation is:

The 1.56 million barrels of “transfers to crude-oil supply” are not new crude production or barrels released from the SPR. They are other hydrocarbon liquids reclassified into the crude stream because they are blended with, or processed alongside, crude oil. For the week, they consisted of 38,000 b/d of Alaskan natural-gas liquids and 185,000 b/d of natural gasoline, condensate and unfinished oils. The EIA adds these barrels to crude supply and subtracts them from other petroleum supply, so this does not represent additional supply for the petroleum system as a whole.

A standout figure is the 3.4 million barrels for “EIA adjustment/timing discrepancy”. This is a balancing item, not an identified physical flow, because the independently collected weekly figures do not balance. The EIA previously called this “unaccounted-for crude oil”. The EIA is effectively saying, ‘Commercial stocks increased by 3.4 million barrels more than can be explained by the reported production, trade, refinery-input and SPR figures.’

There are multiple reasons why this could occur:

• imports understated or assigned to a different week;

• exports overstated or recorded before the crude physically left;

• refinery inputs overstated;

• domestic production understated;

• commercial or SPR stock measurements falling on slightly different reporting cut-offs;

• tanker unloading, pipeline movements and transfers between SPR and commercial facilities straddling the Friday reporting boundary; and

• ordinary survey and estimation error.

The crucial implication is that roughly 20% of the reported 17.42-million-barrel commercial crude build is not explained by observable weekly flows, while another 9% reflects the reclassification of other hydrocarbon liquids into crude supply rather than an increase in total petroleum supply. Together, these account for around 29% of the headline build, although only the 20% balancing adjustment is actually unexplained. The headline is thus considerably less informative than it initially appears. The 17.42-million-barrel build is supported by the commercial-stock survey, but the weekly flow data cannot explain a fifth of it. That strongly supports treating it as a noisy logistics-and-reporting event, not evidence that an oil glut created an unexpected 17.42-million-barrel commercial surplus.

In addition, the SPR draw continued at a high rate, reducing the reserve by another 6.1 million barrels to 298.7 million barrels, its lowest level since 1983. Combining commercial and strategic crude stocks therefore produces a net build of only 11.30 million barrels, rather than the headline 17.42 million. The continuation of an emergency SPR draw is also difficult to reconcile with interpreting one week’s commercial build as evidence that the US has moved into a supply glut.

Further, the identifiable market surplus was created by unusually high imports and low exports during one week, rather than necessarily indicating a durable change in supply. The high net imports overcame the structural market deficit. It is structural because many complex Gulf Coast and Midwest refineries are configured for Canadian, Venezuelan and other imported heavy or medium-sour grades, while the local production is mainly light crude, which is either exported or blended with the heavy for domestic refining.

If the current week’s production and refinery runs are retained, but imports and exports are replaced by the previous week’s figures, the calculation instead produces the following crude deficit.

The week-on-week increase in net imports therefore improved the balance by 1.767 mb/d, or 12.37 million barrels over the week. It converted what would otherwise have been a 6.03-million-barrel deficit into a 6.35-million-barrel identifiable surplus.

Nor did the crude build translate into broad petroleum-product replenishment. Gasoline inventories fell by about 1.0 million barrels, distillates were essentially unchanged and remained around 12% below their five-year average, while total commercial petroleum inventories increased by only 15.7 million barrels, less than the crude build alone. This is difficult to reconcile with the idea that the US petroleum system is moving into a general glut.

So, does this US commercial crude inventory build tell us anything? Not really very much by itself. It mostly means that an unusually large amount of crude arrived in the US during one particular reporting week while exports fell, and refineries – already running at 96.2% – could not absorb the extra barrels. Part of the reported build also came from the SPR, part reflected the reclassification of other hydrocarbon liquids into crude supply, and one-fifth was the EIA’s balancing adjustment.

Rather, it provides information – not totally unexpected – about the current US logistics and trade balance. Canada and Venezuela are still supplying crude, delayed overseas cargoes are arriving, US crude exports have weakened, and the US Gulf Coast temporarily received more crude than it could process or ship onward. There is quite a bit that it does not tell us, even though the data may be used for that misleading purpose. It does not yet demonstrate a global crude surplus, meaningful recovery in Persian Gulf supply, weakening refinery demand, replenishment of the overall US buffer, or an end to refined-product tightness.

So where are we now? The short version is that Iran and Oman appear to have agreed on where ships could travel through the Strait of Hormuz, but almost nothing has been publicly agreed about the conditions under which they would travel.

Early in the week, it sounded as though a deal might be announced almost immediately. US Treasury Secretary Scott Bessent said Hormuz could reopen within a day or two, Trump said negotiations were moving along nicely, and Qatar confirmed that indirect US-Iran talks had reached “very progressive stages”, with draft language being circulated between the parties. Unsurprisingly, oil responded accordingly as traders unquestioningly accepted these assurances yet again, with Brent falling below $80 a barrel and WTI down near $75.

Then – like “déjà vu all over again” – the proposed terms began to emerge.

Iran and Oman appear to have agreed on the coordinates of a temporary route through the strait, whereby vessels entering the Persian Gulf would travel through Iranian waters under Iranian management and those leaving would initially use mostly Omani waters. Later Iranian accounts amended the latter to describe outbound traffic as being jointly overseen by Iran and Oman. Eventually, both existing routes would be replaced by a central corridor administered under the new arrangement. This temporary route is to operate while negotiations continued over a permanent settlement. The initial period seems to be 60 days, although Iranian officials have suggested it could continue for as long as four months. Iran is also reportedly considering allowing European countries to clear mines from the waterway, which would provide shipowners and insurers with at least some assurance that it was safe.

Now all of that sounds quite practical, which is somewhat unusual for these negotiations. After months of competing American and Iranian routes, ships would have one agreed channel and a common process for moving through it. The ambiguity over navigation that helped destroy the June ceasefire would finally be resolved. Except the route is the easy part. The dispute is over who controls it, who may use it and whether Iran can charge for passage.

Iran wants to bar US and Israeli vessels, restrict cargo linked to Israel, require compensation from countries it considers responsible for the war and charge for services associated with transit. Reports state that Iran is seeking fees equal to 5-7% of cargo value, while Oman has considered a charge of around 3%. These are not modest port or pilotage fees. A VLCC carrying $150 million of crude would face an Iranian charge of $7.5 to 10.5 million. Additionally, payments to Iran’s Persian Gulf Strait Authority could breach US sanctions, exposing shipowners, banks and other intermediaries to penalties or frozen assets. New Lloyd’s Market Association clauses apparently terminate cover if a vessel pays the disputed fees. A tanker owner, therefore, would face the risk of complying with Iranian rules, violating American sanctions and losing insurance coverage. Clearly that is not a workable basis for commercial shipping.

Unsurprisingly Washington has rejected this entire concept. A US official said on Thursday that any temporary route must operate without approvals, permissions, tolls or charges. The US wants a return to something resembling the pre-war position, under which Hormuz was treated as a waterway open to international shipping rather than one administered by Iran.

The situation is therefore that Iran and Oman appear to have agreed on a route, but the terms Tehran attaches to it remain unacceptable to Washington and potentially unworkable for commercial vessels. So, in other words, we remain at an impasse over the same issues, and the market wearily moved Brent back up to $83.55 and WTI to $78.18.

Iran is not offering the old Hormuz back

The easiest way to misunderstand these talks is to assume that Iran is negotiating the reopening of the pre-war Strait of Hormuz, but it is not. Tehran is offering a new arrangement built around the leverage it has acquired during the war; managing ships entering the Persian Gulf, retaining visibility or joint authority over ships leaving, excluding vessels or cargo it considers hostile, and receiving compensation or fees for allowing passage. One person familiar with the talks told Reuters that the concession has already been made on some form of Iranian control over the strait, whatever form the final language takes, so Washington accepting Iranian control without acknowledging it explicitly appears to be the likely outcome here.

Iranian legislators are considering a bill that would ban US and Israeli ships, restrict cargo connected with countries supporting Israel and fine violators as much as 20% of cargo value. The legislation may never take effect and is probably intended partly to strengthen Iran’s bargaining position, but it reveals how Tehran now views Hormuz. More than just as a waterway beside its coast, it is an instrument through which it can regulate the economic relationships of other countries.

The US position is almost the reverse. It insists that Iran has no right to decide which commercial vessels may transit an international strait and no right to charge them. Any formal recognition of Iranian control would, Washington argues, undermine freedom of navigation and establish a precedent that could be applied at other maritime chokepoints. Yet, as has been clear for months, the superpower does not have the sway it thought it had. This is why what had looked like a promising agreement may be another false dawn. Agreeing the route coordinates is not the same as agreeing to reopen. Iran and Oman have apparently decided where a ship can sail, but the US and Iran have not decided what sailing through that route would mean.

The discussions are raising more questions than they answer. Would a transiting vessel need explicit Iranian permission? Could Iran inspect its ownership or destination? Could it prohibit cargo bound for Israel? Would shipowners pay Iran, Oman or a jointly administered body? Would the US waive sanctions on those payments? Would insurers continue covering the vessel? And what happens when Iran classifies a ship as hostile but Oman does not? Until those and other questions are answered, the central corridor is simply a line on a chart rather than an open trade route.

The price of reopening is lifting the blockade

The other essential part of any agreement is the US blockade of Iranian ports. Iran has not promised that agreement with Oman would automatically reopen Hormuz – it is a necessary but not sufficient requirement. Tehran says the US naval blockade of Iranian ports must first be lifted. CENTCOM has interdicted or redirected dozens of ships since restoring the blockade in July, preventing Iran from moving its oil and other trade while insisting that Iran allow everybody else’s cargo through the strait. From Iran’s perspective, reopening Hormuz while its ports remain blockaded would amount to surrendering its principal leverage for nothing. It would restore the oil exports of US partners, relieve pressure on the global economy and reduce the political cost of the war for Trump while leaving Iran economically strangled. Obviously not a trade Tehran has any reason to accept.

The June Memorandum of Understanding (MOU) explicitly recognised this connection. The US agreed to lift the blockade, restore waivers for Iranian oil exports and begin releasing frozen Iranian assets, while Iran agreed to permit free passage through Hormuz and maintain the nuclear status quo during a 60-day negotiating period. That MOU broke down, however, because the sections dealing with Hormuz were too vague. Washington believed ships could use the southern route without Iranian involvement while Tehran continued to insist that vessels follow authorised channels and accept some Iranian oversight. Ships were attacked, the US resumed strikes and sanctions, and the blockade was reimposed. This is a recognised problem with attempting to negotiate international agreements in brief and in a hurry. It is well known in diplomatic circles that such documents are complex and liable to misinterpretation, especially if they are in more than one language. This is why they normally take months or even years to finalise, and frequently run to dozens or hundreds of pages. The MOU attempted to compress a ceasefire, blockade relief, maritime administration, sanctions waivers, nuclear restraint and a pathway to a permanent settlement into just 14 brief points, while leaving the most disputed operational terms undefined. It was more an agenda for further negotiation than an agreement, so had little chance of lasting.

Washington had previously not indicated that it was willing to restore the other half of the June bargain, but that possibly changed on Friday. A US official said Washington expected an Iran-Oman agreement to be announced soon and would lift its blockade of Iranian ports if Iran then provided unimpeded commercial passage through Hormuz. This is the first clear indication that the Oman route could become part of another US-Iran ceasefire rather than remain a bilateral arrangement that Washington rejects, because it provides the necessary exchange whereby Iran restores commercial passage and the US removes the blockade. Yet the sequencing problem remains. Washington says lifting the blockade will be “performance-based” and conditional on Iran implementing the agreement. Iran has maintained that it cannot fully reopen Hormuz while its ports remain blockaded. The US therefore wants Iran to surrender its principal leverage first and promises relief once it is satisfied with Iranian performance. Tehran wants the blockade lifted as part of the opening itself. And still the devil remains in the detail. The parties have not publicly agreed on what unimpeded passage actually means. For Washington, it means no Iranian permissions, discriminatory restrictions or tolls. For Tehran, the emerging arrangement appears to preserve Iranian management of inbound traffic, restrictions on hostile vessels, and charges for maritime services.

Are the US and Iran negotiating?

There are negotiations between Iran and the US, but not in the way Trump has implied. Iran initially insisted that it was holding talks only with Oman. Trump responded by calling the Iranian leadership “unbelievably duplicitous”, claiming that Tehran had requested negotiations while publicly denying they existed. Qatar subsequently confirmed that indirect US-Iran talks were under way and that draft language was circulating. Qatar and Pakistan are reportedly working with Oman, while Saudi Arabia and Iraq have also carried messages and pressed the parties to avoid another escalation. So there is an active diplomatic process, but apparently no direct negotiation between American and Iranian officials. Messages are moving through several intermediaries, allowing Iran to talk to Washington through backchannels without publicly appearing to negotiate while under bombardment and blockade.

This is more than a semantic distinction or simple posturing. Iran needs to present any agreement as the product of regional diplomacy rather than capitulation to American threats. Trump, meanwhile, needs to present it as a deal he personally imposed upon Iran. The art of successful negotiation, especially in the diplomatic arena, is to arrive at a position such that both sides can present their preferred – yet, ostensibly mutually exclusive – cases to their respective audiences at home while coming to an agreement through compromise. This requires understanding your opponent intimately, a failing that the US has demonstrated numerous times over the years, especially with respect to Iran.

Whether Iran can approve a final document at all is a separate obstacle. Any agreement would require the consent of Supreme Leader Mojtaba Khamenei, who has remained in hiding since being injured during the initial US-Israeli attacks, and President Masoud Pezeshkian says communication with him is currently “very difficult”. The diplomats may therefore be circulating language without knowing whether Iran’s ultimate decision-maker has accepted it, or will do so. That is a more realistic and meaningful obstacle than JD Vance’s convenient and simplistic division of the Iranian state into people who want peace and “crazy radicals” who do not.

The US has run out of useful military options

Trump keeps threatening enormous attacks and then postponing them. This is partly posturing and partly a recognition that he is now aware that he has limited military options. The US has demonstrated that it can destroy Iranian civilian and military infrastructure on a huge scale. It has reportedly struck more than 13,000 targets, damaging air bases, missile sites, ports, industrial facilities and government buildings. Iran’s economy has suffered losses measured in the hundreds of billions of dollars, inflation has surged and perhaps a fifth of its industrial base has been damaged. Yet the bombing has not forced Iran to capitulate or even to reopen Hormuz. The historical record offers little support for the idea that bombing alone reliably forces a determined state to surrender. NATO’s 1999 campaign in Yugoslavia is the closest modern example, but even there the outcome followed 78 days of bombing alongside Russian diplomatic pressure, growing isolation and a credible threat of ground intervention. Iran is much larger, retains greater retaliatory capacity and is fighting over what its leadership regards as an existential issue.

On the naval front, Iran does not need to defeat the US Navy in a conventional battle. It needs only to retain enough missiles, drones, mines and small boats to make a few tanker owners believe that their vessels may be hit. Trump acknowledged this himself on Thursday: the US may claim to control the waterway, but Iran “can always shoot something”, and owners do not want to risk valuable ships. That sums up the whole war, and why a short “excursion” has turned into a 5-month stalemate. Military control of the sea is not the same as commercially reliable passage through it. Even with a navy not totally suited to the task, the US can still escort ships, patrol the Omani route and strike Iranian launch sites, but it cannot guarantee that a tanker will not hit a mine or be attacked from the Iranian coast. If crews refuse to sail, insurers withdraw cover and shipowners keep vessels outside the Gulf, then any declarations of American control have little practical value. ADNOC, the UAE’s national oil company spent $1.3 billion this week to nearly double its VLCC fleet from eight to fourteen, bringing more of the transport chain under its own control, although this does not remove the underlying missile, mine and insurance risks.

It was reported that a senior officer in CENTCOM’s intelligence branch emailed a broad group of military analysts asking for “new creative and unconventional ways to pressure and punish Iran”. This is revealing as it shows that after five months of fighting, the military is out of ideas and looking for new ones because further iterations of the existing campaign will not produce the result Washington wants. Trump’s recent threat of the “largest attack since the Second World War” was withdrawn after Gulf leaders urged him to negotiate. Reports that shortages of missiles and air-defence systems influenced the decision have been angrily denied by Trump, even as the Washington Post reported he privately confronted Defense Secretary Pete Hegseth over having been misled about the state of those stocks. Regardless of the denial, it remains a fact that the heavy depletion of interceptor stocks is a significant problem, especially while Iran retains enough inexpensive offensive weapons to continue imposing costs. Although the US can inflict much greater destruction, Iran has nevertheless retained the cheaper path to achieving its immediate objective. It is reminiscent of past American wars, like Vietnam.

Iran is paying dearly too

Of course, none of the above means that Iran is winning without incurring meaningful costs. Its military and industrial infrastructure has been badly damaged. The rial has fallen, inflation has risen dramatically, unemployment has increased and the naval blockade is choking trade. Estimates of total losses vary enormously, but even the lower figures are severe. Rebuilding the petrochemical complexes at Mahshahr and Assaluyeh could take more than decade. Pezeshkian has described the country as facing its most difficult period since the Islamic Revolution. A permanent war economy built around shadow shipping and retaliation is not a viable development strategy, so Iran will eventually need foreign currency from oil exports, access to its frozen assets and a credible route towards reconstruction.

Yet suffering does not automatically produce surrender. Iran has lived under Western sanctions for decades, has developed methods of evading them and can make the consequences of its economic hardships felt far beyond its borders. More importantly, foreign attack also strengthens the hardliners’ argument that compromise with Washington is impossible and encourages at least some Iranians who had been opposing the Government to rally around the country against an external aggressor. Iran’s long history of foreign intervention gives that response deep political resonance.

Strategy and the Countdown Clocks

The strategic question is therefore not whether Iran is being hurt, since it obviously is. The question is whether the pain being inflicted changes Iranian decisions before its retaliation changes American decisions. So far, it has not.

It is a contest between two clocks. One clock measures Iran’s economic deterioration. The blockade is suppressing exports, industrial damage is accumulating and the cost of eventual reconstruction rises every day. The other clock measures the depletion of the world’s ability to absorb the disruption: commercial inventories, strategic reserves, missile interceptors, political patience and household tolerance for higher fuel prices. Iran is suffering more in absolute terms, but Trump faces the shorter political timetable. The US midterm elections are in November, petrol prices remain well above their pre-war level, and the administration needs lower energy prices quickly. Iran does not need a healthy economy by November – it needs only to avoid capitulating before the political cost becomes unacceptable in Washington.

The brief June ceasefire made that contest still more difficult for the US. Iran received temporary sanctions relief and used the window to export stored oil and generate several billion dollars in foreign currency. Oil prices fell, encouraging the US to believe that the crisis had passed, but global inventories and strategic reserves did not fully recover. When fighting resumed, Iran entered the next round with some additional funds while the global buffer remained thin. The threat of more bombing is less persuasive than it may appear. Iran knows that another major escalation would send oil prices higher, accelerate inventory depletion and hurt Trump before it necessarily forces Tehran to concede.

The clock does not have to run out completely. It only has to convince Trump that accepting an unattractive agreement is preferable to continuing the war. And that is what you would expect – Iranian patience outlasting that of the West. An authoritarian regime and patriotic populace facing an existential threat against liberal democracies undertaking a war of choice. Iran’s leadership can suppress dissent, absorb severe domestic hardship and portray the war as resistance to an existential foreign attack. Trump, by contrast, is fighting a discretionary war while facing elections, higher fuel prices and declining public support. The two clocks are not running at the same pace.

The war is no longer confined to Hormuz

The opening of a second maritime front has made that determination more urgent. Saudi Arabia had responded to the closure of Hormuz by sending more crude through its East-West pipeline to Yanbu on the Red Sea, but the Houthis have now declared a blockade of Saudi ports, attacked tankers and Aramco facilities, and expanded their strikes against Saudi-aligned forces in Yemen. Riyadh is trying to contain the escalation through Omani mediation rather than return to the unsuccessful bombing campaign it conducted in Yemen before the 2022 truce. The US is also discouraging a major Saudi response because widening the conflict around Bab el-Mandeb would threaten the remaining alternative route for Gulf oil.

Regional security arrangements are adjusting under the same pressure, separately from anything involving Hormuz directly. Saudi Arabia, Turkey and Pakistan signed a mutual defence pact in Mecca on Friday, agreeing that an attack on any one of the three would be treated as an attack on all three. It is difficult to view that without reference to the context of the past five months during which US security guarantees in the region have looked less reliable than they did pre-war. Three major regional powers choosing to underwrite each other’s security while Washington negotiates the end of its war certainly appears to reflect the diminished sway Washington still has left.

Some Saudi crude is still leaving through Yanbu. Tankers have crossed Bab el-Mandeb with their transponders switched off, while others have travelled north through Suez or taken the longer route around Africa. However, these movements are irregular, expensive and exposed to further Houthi attacks. The disruption is visible in trade flow data, with US imports of Saudi crude falling to zero for the entire month of July – the first month without a Saudi delivery since 1985 – after exceeding 0.8 mb/d earlier in the year. American refiners have substituted domestic light crude and Venezuelan barrels where possible, despite differences in crude quality and refinery configuration. Although US imports from the region may rise again in August as cargoes loaded during the June reopening finally arrive and Saudi barrels are redirected through Yanbu and Suez, that still does not change what July demonstrated.

Traffic has fallen further while the parties negotiate. Only 33 commodity vessels crossed Hormuz between Monday and Thursday, down from 50 over the same period a week earlier and around 130-140 a week before the war. Only four vessels in total crossed on Thursday. More importantly, 21 crude tankers entered the Gulf during the week, but only six exited. That imbalance suggests that Iran is continuing to allow vessels into the Gulf but restricting their departure.

Iran has also warned Gulf Governments that their oil fields, refineries, electricity grids, water infrastructure and transport networks could be attacked if they fail to persuade Trump to end US strikes. That places Saudi Arabia, Qatar, the UAE and other US partners in an impossible position. They depend upon American security support, but American escalation makes them the most accessible targets for Iranian retaliation. Their pressure on Trump to cancel the planned attack indicates that they increasingly regard continued US escalation as a threat to their security.

The oil market believed the headline… again

The movement in oil prices over the past week shows a paradoxical situation, illustrating both a wilful disregard of the fundamentals – preferring to believe a solution is nigh and oil prices will fall – then suddenly switching to a demonstrated lack of confidence in the process as oil prices rise again. Brent fell below $80 when Bessent suggested that an agreement could reopen Hormuz within days. However, once the Iranian terms became clearer and reports emerged of Iranian strikes on targets described as hostile near Qeshm Island, it rose almost 4% in a single session and then continued above $83.

The futures price, too, should not be mistaken for a complete measure of the shortage. Prompt physical crude has continued trading above the paper benchmark because refiners must pay for barrels that can actually be delivered to them, not futures contracts reflecting expectations about a possible agreement. A fall in Brent on optimistic headlines does not result in stranded Gulf barrels suddenly appearing at a refinery. The same issue applies to any reopening. A diplomatic announcement will produce an immediate fall in futures, but restoring physical supply will require mine clearance, sanctions waivers, insurance, willing crews, available tankers and evidence that the arrangement will last past its first disputed transit. Even a fully signed agreement would therefore restore the roughly 20 mb/d pre-war flow only gradually rather than overnight.

What would an actual agreement require?

At this point the terms needed for agreement are reasonably obvious, however difficult it may be for either side to admit. Iran would have to stop attacking commercial shipping, abandon discriminatory restrictions on cargo, accept a meaningful limit on its ability to obstruct passage, and maintain the existing freeze on its nuclear programme. The US would have to end its attacks, lift the naval blockade, restore authorisation for Iranian oil exports, waive sanctions on whatever administrative mechanism manages the strait, and provide access to at least some frozen Iranian assets. The June MOU already contained most of this bargain, but lacked clarity and detail, so the current talks are predominantly about trying to make the old agreement operational and politically presentable.

Washington has now offered to lift the blockade if Iran first provides unimpeded passage, so the outlines of another temporary agreement finally exist. Yet, the same questions which destroyed the June MOU remain unresolved: who moves first, what constitutes free passage, and how much authority Iran retains. Washington means passage without permissions, discrimination or tolls, and Tehran wants its management role recognised and the ability to restrict hostile vessels and charge for services. Both sides also need to claim victory. Trump wants to say that military pressure reopened Hormuz and froze Iran’s nuclear programme. Tehran wants to say that it survived the assault, ended the blockade, restored its oil exports and secured recognition of its role in the strait. A regional management arrangement may contain enough ambiguity for both accounts, although that same ambiguity could cause it to unravel at the first disputed transit.

There is already a precedent for that dynamic breaking a deal rather than completing one. The first round of high-profile talks after the June MOU was signed fell apart because Trump took offence at comparisons between the MOU’s terms and Obama’s 2015 nuclear deal. He threatened to attack Iran and take it over, and the Iranian delegation walked out. The need to look like the side that won has already wrecked one round of this negotiation, and nothing about the current fee dispute makes that need any smaller.

The available settlement is therefore much narrower than the outcome Trump promised when he began the war. Further bombing would damage Iran but would not remove its ability to disrupt Hormuz. Instead, it would provoke more attacks, raise energy costs and leave the same negotiation waiting at the end. The war has reduced America’s objective from forcing Iranian capitulation to finding a tolerable description of compromise. An agreement now depends upon whether Trump can accept that result – and continue accepting it once his critics describe it accurately.

A new moral outrage is working its way around X, and as usual it misunderstands the basics. The trigger this time is Project Panama, Anthropic’s programme to buy millions of second-hand books, cut them from their bindings, scan the pages, and recycle the paper. Court documents unsealed in January provided the name and the ambition – an “effort to destructively scan all the books in the world” – and the viral response went straight to Fahrenheit 451, Nineteen Eighty-Four, and the Library of Alexandria. These comparisons need a closer look than the cursory invocation they are getting.

The programme destroyed one purchased print copy for every digital copy it created, removed nothing from public access, erased no text, and paid the market price for everything it consumed. The books that went into the cutters were the surplus of a second-hand trade that pulps millions of volumes every year without a word of protest. There is a scandal here, but it is the one that very few are posting about: a copyright regime that locks commercially dead books away for a century, so that the scans exist but nobody may read them.

Two programmes, one settlement

Some accuracy first, because part of what Anthropic did was indefensible and has been paid for. In 2021 and 2022 the company downloaded more than seven million books from pirate libraries – Books3, Library Genesis, and Pirate Library Mirror – to build a permanent research collection. Judge William Alsup rejected fair use as a defence for a library built from pirated files, and Anthropic settled for $1.5 billion, roughly $3,000 for each of the 482,460 works in the settlement class and the largest copyright recovery in history. A federal court granted final approval this month.

The scanning programme was the lawful successor to the piracy, and its whole design addressed the previous legal failure. In February 2024 Anthropic hired Tom Turvey, formerly head of partnerships for Google Books, and began buying used books in bulk from wholesalers such as Better World Books and World of Books. Hydraulic cutters removed the spines, high-speed scanners digitised the pages, the paper went to recycling, and the company retained one internal digital copy in place of each destroyed original. Alsup ruled in June 2025 that this was fair use.

The viral posts conflate the two programmes, and their arguments depend on the confusion. “They paid $3,000 per shredded book” divides the piracy settlement by the wrong denominator and attaches it to the wrong programme: the $1.5 billion resolved the piracy claims, whereas the purchased books generated no damages because the court held the scanning to be fair use.

You don’t buy rare books by the pallet

Consider first the Library of Alexandria, since that comparison carries the heaviest emotional charge. The popular telling – sole copies, everything lost in one fire – simplifies a decline that ran across centuries, but the horror it preserves is accurate in one respect: texts died there that existed nowhere else, which is why the loss still stings two millennia later. Project Panama drew its stock, however, from the opposite kind of market – used-book wholesalers selling by the pallet, at cents per volume and from an inventory the world didn't want. The phrase “valuable rare books by the pallet” is an oxymoron, since valuable rare books trade individually, through dealers and auction houses, each catalogued with its history of ownership, at prices that would make feeding one into a hydraulic cutter an act of commercial self-harm. A wholesaler who put a valuable book on a bulk pallet would be destroying their own margin.

The pallets overwhelmingly contain the stock the second-hand trade cannot otherwise sell, such as superseded textbook editions, obsolete software manuals, tax guides for abolished taxes, and the annually revised handbooks whose 1993 edition became waste paper in 1994. These titles meet the technical definition of “rare” – out of print and few surviving copies – but they are rare because millions of copies were pulped, donated, and landfilled over decades while nobody objected. Publishers pulp unsold stock as routine and libraries weed their shelves into skips every year, yet the outrage mourns the books Anthropic is destroying as though they were antiquarian treasures when charity shops refuse to take them.

The provenance of the rare-books alarm needs scrutiny too, because its source is different reporting to the Project Panama documents and relies on far thinner evidence. The court documents that describe Anthropic’s programme came through the Washington Post in January, whereas the “rare books being destroyed” wave traces to a single 404 Media investigation this month, and its named ‘evidence’ consists of two things. First, the marketing copy of ISBNdb, a data broker that arranges bulk purchases for AI labs and advertises that “the world’s best AI training data is sitting on a shelf”, and second, anonymous booksellers who reported surges in bulk ISBN-based orders from buyers they could not identify. The sellers suspect AI labs, but they do not know. The anxiety about rare texts thus comes only from a vendor’s sales pitch and inference from order patterns, yet every subsequent headline recycles that one report. There is still no documented case of a rare antiquarian book destroyed – the commentary describes worry about what might happen, but worry is not evidence.

Why the AI labs want these books in the first place clarifies the issue: they prize the date rather than the scarcity. Print published before 2022 predates the flood of machine-generated text that now contaminates the open web and degrades models trained on it, and a purchased physical book gives the AI lab a documented chain of lawful acquisition. The ideal purchase is therefore a cheap book whose text has never been digitised – scarce in the datasets and worthless on the shelf – since rarity adds cost and legal risk while adding nothing that a common undigitised book does not supply. One bookseller conceded to the same 404 Media investigation that the bulk orders were clearing dead inventory.

Against all of this is the simple observation that an AI company needs only one copy of each title. One copy, drawn from a surplus market, has no measurable effect on circulation or on the availability of information to anyone. Every other copy in the world stays where it is, and if you so desire you can still buy your own copy. That you hadn’t bought one beforehand suggests that it’s not really about the books themselves, but rather the opportunity to express outrage at AI.

The loophole that isn’t

One comment repeated frequently is that Anthropic’s programme is exploiting a legal “loophole”, but that claim inverts the legal logic. The court held that scanning a purchased print copy and discarding the original is just format-shifting: one copy replaced by one copy, with the total unchanged. It is, in everyday practice, the same principle that lets you rip a CD you bought. For a private individual the destruction would be unnecessary, since nobody litigates a personal backup, but a company digitising millions of books commercially needs stricter discipline – the one-for-one replacement kept the programme inside the doctrine, because retaining both copies would have multiplied them and weakened the case that this was replacement rather than reproduction. The destruction thus made industrial-scale scanning defensible rather than bypassing something otherwise illegal.

A related objection holds that the destruction flows from the decision to take without consent – that had Anthropic sought permission, no book need have died. That causation fails twice. Anthropic did seek permission, sending Turvey to the publishers to discuss licensing, but the publishers declined, so the counterfactual was actually tried. What remained was the action that copyright has always allowed for, which is purchase. Buying a copy confers ownership of that copy rather than a licence to reproduce the work, but nobody needs an author’s permission to read a purchased book or to learn from it, and the court held that converting the copy to another format for internal use fell within fair use. The destruction is therefore not a consequence of an unlicensed taking but ensuring the conversion was a replacement rather than a multiplication – one copy in, one copy out. The demand for consent-to-train was a demand for a new right, one the statute doesn’t address, and Judge Alsup declined to invent it.

What happens once the model has read the books? Nothing, nor should it. Copyright restrains copying rather than learning, so a model absorbs a lawfully acquired book similar to a human borrower absorbing a library copy, and what a reader takes away and later uses in their own writing has always been beyond copyright’s reach. Alsup made the comparison himself, observing that objecting to training on lawfully obtained books resembles objecting that teaching schoolchildren to write will produce authors who compete with you.

The remaining claims

The rest of the various claims I have seen on X can be dealt with briskly.

“They planned to shred every book on the planet”

They planned to purchase a single copy of every book on the planet. The internal document’s “all the books in the world” describes acquisition of titles, and the difference between one copy of each book and every copy of every book is the difference between a library programme and an act of cultural annihilation.

“The scans are locked in Anthropic’s private library forever”

Anthropic bought the physical copies and retained one internal digital copy in place of each destroyed original, which is the format change the court held lawful. Nothing was removed from public access.

“The authors got a shredded book and nothing else”

A second-hand sale owes the author nothing – no royalty has ever attached to a used book changing hands. Authors and other rights-holders of the pirated works, meanwhile, shared a $1.5 billion payout.

“A book on a shelf lasts 500 years without power, wifi, or a subscription”

A book lasts as long as someone chooses to keep it, and most are not kept – publishers pulp unsold print runs, libraries weed their stacks, and estates clear houses, so millions of volumes go to the skip every year. Nor is the medium as durable as claimed: most books in the past century are printed using acidic wood-pulp paper that yellows and embrittles within decades. The durable-artefact romance describes almost no book that has ever been mass printed.

“The AI it was destroyed for gets thrown away in 18 months”

Models are retired, but the scans are not. The library remains available for future training and research, which makes the scan the durable object in this exchange and the model the consumable.

What the dystopias actually warned about

If you think this is Fahrenheit 451, you need to read it again. Bradbury’s firemen burned books to erase what was in them – suppression was the purpose, and the State hunted every copy so that no one could read the text again. Project Panama preserves the contents verbatim, destroys one purchased copy, and leaves every other copy in the world untouched. A regime that burned a single copy of each book while carefully transcribing the contents would make a very strange dystopia. And that’s not what Bradbury wrote.

Nineteen Eighty-Four fails as a comparison for the same reason but in reverse: the “memory hole” existed to erase the record and rewrite it, whereas a verbatim scan preserves the text exactly as written. The more subtle and conspiratorial fear – that once the physical books are gone the digital copies can be quietly altered – borrows Orwell’s mechanism while forgetting his precondition, because alteration works only where the forger holds a monopoly over the record. The Party burned every original so that no copy survived to contradict the rewrite. One purchased copy per scan creates nothing like this scenario. Tampering with an AI company’s archive would falsify nothing beyond its own training data, and the evidence would be in the millions of surviving copies untouched, on library shelves and in private homes.

The novel that comes closest is rarely cited. In Brave New World the old books are locked away rather than hunted – Mustapha Mond keeps his Shakespeare and his Bible in a safe – yet the State doesn’t really need the prohibition because a public amused into indifference would not read them anyway. Huxley’s warning describes a civilisation that lets its literature go for want of readers. The contemporary wholesale book market selling academic texts by the pallet for cents is that civilisation’s clearance sale. Bradbury, for his part, spent his later decades insisting that his villain was actually the audience who stopped reading rather than the State, with the firemen acting only after the public had stopped caring. The dystopias, properly read, indict the market that supplied the pallets rather than the AI company that bought them.

The argument that we should be having

Which returns us to the scandal worth the outrage: the scans exist yet no one may read them. The productive response to Project Panama would be to open that archive rather than stop the machine, but the obstacle is a copyright regime calibrated for a world that no longer exists. One attempt did come close. The Google Books settlement of 2008 would have made out-of-print books viewable and purchasable online, with authors paid through a registry. However, Judge Chin rejected it in 2011, partly for inverting copyright’s opt-in default and partly on the view that no private company should acquire through a class action a licence which Congress had not granted to anyone. The outcome thus combined the worst of every option – the infrastructure was built and the books were scanned, but the archives remain restricted. The public did get snippets – Google Books still serves search results, previews, and public-domain texts, but the full access the settlement promised died with it, and Anthropic’s archive offers no public window at all.

The Internet Archive tried another route. Its public-domain collection is uncontroversial – anything published before 1931 is out of copyright in America and freely readable on the site. However, for in-copyright books it ran controlled digital lending: scan a book you own, lend one digital copy at a time, and withdraw the print copy while the scan is out. So you retain just one copy in circulation, which is similar to the one-for-one replacement logic that Alsup later accepted for Anthropic. It didn’t last because the publishers sued, the Second Circuit ruled against the Archive in 2024, and roughly half a million books came down. The court’s reasoning was that the Archive lending a scan is the same use a publisher’s licensed ebook serves, so every loan potentially displaces a sale. This is different to Anthropic training a model on a scan, as no copy substitutes for a licensed ebook. Instead, the text is subsumed into the model’s statistical weights, and the scan stays in the vault without reaching the public. The first use is thus considered to compete with the book in its own market while the second does not, which is why AI training was deemed acceptable and digital lending was not. Hence the strange position the law has now created: a company may scan a book for a machine’s private consumption, but a library that scans the same book to lend it to one reader at a time loses in court, at least wherever the publisher sells an ebook.

Copyright’s term causes much of the problem. Life plus seventy years keeps a 1993 book by a long-lived author locked until near the end of this century, decades after its commercial life ended, and for a large share of out-of-print titles the rights holder cannot even be located to ask. The incentive function of copyright is exhausted within a decade or two of publication for most books, and everything after that is deadweight. Several European states have a workable alternative – collective licensing for out-of-commerce works, under which a collecting society authorises digitisation unless the author objects. That preserves the opt-out Google Books was denied while locating it in statute rather than settlement. If that model were extended to the training archives then the public could gain access to the contents of every book nobody sells any more. The people posting Bradbury quotes should be demanding exactly this, and the fact that none of them are suggests the outrage was directed at the machine because the machine is easier to hate than the law.

The war has paused rather than resolved, yet again, and the difference is starkly illustrated by the flow data. Goldman Sachs estimates reported oil exports through Hormuz at just 1.6 mb/d (million barrels a day), 8% of the pre-war 20 mb/d.

Vessel tracking has recorded no documented outbound crude capacity through the strait on either 26 or 28 July. Yet, as has been the way of this event, oil has swung with the diplomacy rather than the physical picture. Brent fell from last Thursday’s intraday high above $100 to around $84 by Monday morning. It edged up to roughly $84.50 Monday evening after Iran’s missile attack on US forces, while WTI moved more, from below $78 to about $83.

As with previous pauses, this one has not seen a clean respite in strikes. Iran launched ballistic missiles at American forces in Jordan on 28 July, all intercepted according to CENTCOM. The launch came hours after Trump met Israeli Prime Minister Netanyahu at the White House, and after Trump had told Axios he was “considering a massive attack, bigger than ever before… we are all set for it”, while separately repeating the bridges threat – “I could take out most bridges in less than an hour”. CENTCOM also disclosed that Iran-aligned militias have attempted more than 600 attacks on US citizens and facilities in Iraq between February and April alone, so this Jordan strike is one incident in a much larger campaign, rather than an isolated one. The US hadn’t resumed its bombing campaign as this was being written, making this an armed pause rather than a second ceasefire. Trump says talks are under way, but Iran says they aren’t. Messages are moving through Oman, but the underlying dispute has changed from whether Hormuz reopens to who controls it once it does.

The news that is circulating regarding diplomacy overstates what is actually confirmed. Reuters and Al Jazeera both report, citing Iran’s deputy foreign minister Kazem Gharibabadi, that Oman proposed a Malacca-style arrangement – equal transit lanes, voluntary rather than compulsory fees – and that Iran rejected it, on the stated grounds that equal division “doesn’t address Iran’s security concerns”. Iran’s counter-proposal was one shipping lane entirely within Iranian territorial waters, part of the other lane also under Iranian control, and no third country permitted to conduct mine clearance even at Oman’s invitation. This rejection is direct from Gharibabadi himself rather than from anonymous officials, but some reporting has gone further, describing unnamed mediators as “incensed” and casting Iran’s counter-proposal as grabbing “the lion’s share” of the strait. Neither Reuters nor Al Jazeera use language anywhere near that strong, thus it reads more like editorial colour from a source with an interest in making Iran look like the unreasonable party. It is also useful to note that the current round of Iranian attacks traces back to Oman, which had set up a US-backed shipping corridor along its own coastline without Iranian approval, prompting Tehran to resume strikes on shipping and the US to respond in kind. Gharibabadi also directly contradicted Trump’s public account of where things stand – Iran has not submitted a single request to negotiate with Washington in the past fifteen days, despite Trump repeatedly telling reporters that Iran wants to talk.

The Red Sea vulnerability flagged previously is now apparent in the data. Yanbu port loadings have fallen to 3.3 mb/d on a seven-day average, down from 4.3 mb/d over the trailing month, and empty tanker capacity inside the Red Sea and Gulf of Aden has dropped 15% since the Houthis announced their blockade. The bypass that had been absorbing a large share of the Hormuz shortfall is thus tightening under the weight of the second front.

China’s position has been highlighted several times previously as the major swing factor, and there are early signs it is adjusting its stance. Net crude imports are up 3.4 mb/d over the past fortnight, though still 3 mb/d below levels a year ago, and refinery utilisation climbed to 66% in July. That is a partial rebound, but the barrels recorded as arriving now were probably bought weeks earlier, when prices were lower.

Uncertainty remains, nonetheless, over two related variables: whether China keeps buying at current prices and how large China’s pre-war cushion actually was in the first place. Official Chinese data implied a stockbuild near 1.4 mb/d in the year before the war, with a cumulative build since 2017 approaching 3 billion barrels. Satellite-tracked estimates, however, put the cumulative observed increase since 2017 far lower – closer to 450 mb at its peak, roughly one-sixth of the build implied by official data. In the year immediately before the war, those observed stocks rose by around 0.375 mb/d, just over a quarter of the official implied rate of 1.4 mb/d. That suggests the additional pre-war cushion was substantially smaller than the official flow data implies.

The SPR (US Strategic Petroleum Reserve) debate is complex, needing a more considered view than “Washington is bluffing”, and so a full treatment will be considered in a separate piece. Two things are worth raising now, however.

First, DOE (US Department of Energy) has already signed contracts reducing the reserve to roughly 282 mb once current deliveries finish. This is below the 300 mb some analysts treat as a hard floor. If that was the real floor, committing to deliveries past it could expose DOE to legal risk, so the contracts themselves are evidence that it is not the floor.

Second, DOE’s claim that the real floor is only 70 mb also has more behind it than a talking point invented for this crisis. I was sceptical of DOE’s 70 mb figure last week, mostly on incentives – no government understates its reserves without a reason. Although the new evidence doesn’t fully settle that either way, it does complicate it. A 2016 DOE report said each site must be able to draw down until 90% of its inventory is depleted and the same standard was repeated in DOE’s most recent report this year. It works out that 10% of the SPR’s 714 mb capacity is almost exactly 70 mb. That’s a decade-old paper trail, not a recent invention. Whether 70 mb is the real floor, or whether it is higher, depends on unknown technical factors, such as how much oil the Gulf Coast pipelines can actually carry rather than how much is in the caverns. For now, the observation I made last week looks weaker than it did then, although the reasoning behind it still holds until demonstrated otherwise. Still, the more interesting and crucial constraint may not be how much oil is left underground at all.

Positioning data in the futures market backs up the buy-the-rumour-sell-the-fact reversal already made above. In early July traders were holding near their lowest net long position on Brent and WTI combined in years – these were bets for a calm market, not an escalating one. This setup means that if Chinese buying picks up or the coordinated SPR releases slow, all those traders would have to buy back their positions at once – which is exactly how a short squeeze starts.

Not everyone reads Trump’s decision to hold off on renewed strikes as good news, either. One argument is that Iran is negotiating from a position of confidence, not weakness, and has little reason to make the concessions that a complete resolution needs. If that is right, the market has rushed to price in peace at least three times already this year, wrongly each time. This may be yet another.

The two-spikes thesis remains live. Hormuz’s reported oil-export flow is running at roughly one-twelfth of its pre-war level, the Red Sea bypass is now tightening rather than absorbing the shortfall cleanly, China’s suppressed imports may be returning, and the reserve debate continues to be uncertain yet more nuanced than headline numbers suggest. The thesis does not require strikes to resume; rather, it just needs the situation to remain unresolved since those conditions together still point toward a crude supply shock.

Twelve consecutive nights of American strikes have now passed, a second front has opened in the Red Sea with actual attacks rather than just threats, and Trump has turned the bridges-and-power-plants threat into a standing policy. For every future Iranian strike on a ship in the strait, he wrote on Truth Social, the US will “bomb and destroy ONE BRIDGE OR POWER PLANT”, including targets in or near Tehran itself. A specific new target has also been named: Pickaxe Mountain, a facility south of Tehran that Israeli intelligence believes holds buried centrifuges and enriched uranium, with tunnels reportedly a hundred metres underground. Trump says the US will hit it “pretty soon”.

The Red Sea had been a conditional threat – Iran told the Houthis to prepare to close Bab el-Mandeb if the US struck Iran’s power grid – and became unconditional on 20 July, when the Houthis declared a naval blockade against Saudi Arabia outright. The Houthis moved from threat to action on 23 July, claiming missile and drone attacks on two Saudi tankers, the Encelia and the Layla, for allegedly violating the blockade. Kpler tracking data shows at least nine ships turning back, pausing or changing course near the Bab el-Mandeb approach after the blockade announcement. Trump’s response has been dismissive, saying he’ll “take care of” the Houthis if the blockade holds, a consistent habit of ignoring risks rather than engaging them.

The Saudi East–West Pipeline to Yanbu has carried far more oil to the coast than before the war. According to Goldman Sachs, pipeline flow has risen from roughly 1 mb/d pre-war to somewhere above 6 mb/d now, after the parallel NGL (natural gas liquids) lines were converted to crude service. That amount is what the pipeline delivers to the coast, though, rather than what actually leaves as exports. Reuters has reported that actual loadings from Yanbu have averaged more than 4.5 mb/d since April. However, that still leaves a gap between what the pipeline delivers to the coast and what is loaded onto ships, and some of the difference may be accumulating in storage rather than reaching buyers, though the numbers available don’t establish how much. What is clearer, though, is the exposure to the Houthi intervention of this substantial Saudi crude bypass moving from the Red Sea toward Asia.

An effective blockade also raises a critical challenge for the US: can the navy hold two maritime fronts without straining its ability to keep escorting shipping through Hormuz and without needing to draw from other naval commands. The limiting factors are both hulls and magazines. The vessels deployed to Hormuz were already mismatched to the job, better suited to open-ocean confrontation rather than blockade and escort duty. Splitting that same undersized and mismatched fleet between Hormuz and the Red Sea is thus a more complex problem than simply covering more distance. The magazine constraint compounds it as every Iranian missile or drone intercepted consumes at least one interceptor. Each of these advanced interceptors is far more expensive than the missiles or drones, and annual American production of them remains well below the expenditure rate of a sustained campaign. Hormuz and the Red Sea draw on different mixes of interceptors – Patriot and THAAD batteries protecting Gulf bases and infrastructure, naval systems like the SM series and ESSM protecting the ships at sea, Arrow largely Israel’s own platform – but they draw from the same broader, finite pool of American and allied interceptor production. So it’s not one shared magazine running down twice as fast, but several overlapping supply chains all being drawn on at once, speeding up the depletion clock.

The market is finally starting to recognise the risks, with Brent climbing to $89 on Monday, its strongest weekly gain since April, dipping briefly on Tuesday on reports of a ten-day ceasefire proposal from regional mediators, but then pushing through $98 by Thursday, as another tanker was struck – a clear sign of how little that mediation report has ended up being worth. Goldman Sachs updated its forecast numbers, putting Brent above $120 in the fourth quarter and averaging $100 through 2027 if Hormuz stays disrupted that long, against their base case of $80 for 4Q26 and $75 for 2027. Goldman also noted that in the five largest oil-supply shocks of the past fifty years, production in the affected country was still down 42% on average four years later, because infrastructure damage and lost investment kept output suppressed long after the fighting that caused it had ended.

Refined product margins have moved further than crude, which reflects two independent drivers. The New York diesel crack – the spread between the finished product and the crude behind it – hit $88 a barrel, the highest on record. Jet fuel was trading at $154 a barrel in Europe, nearly double the cost of the crude it is derived from. Part of that results directly from Iran, through lost Gulf refining capacity. Part of it also comes from somewhere else entirely: according to the Telegraph’s Ambrose Evans-Pritchard, Ukrainian long-range drone strikes have knocked out around 4 mb/d of Russian refining capacity, more than half the country’s total capacity. There’s a paradox in how the Ukrainian strikes affect prices. Crude that Russia can no longer refine domestically doesn’t stay in the ground, but instead is freed up for export, which acts as a mild counterweight to current upward pressure on benchmark crude. However, Russia is banning diesel, petrol, and jet fuel exports outright to protect what’s left for its own market. Evans-Pritchard puts the combined loss from the Gulf and Russia together, plus what’s potentially at risk from a Houthi closure of the Red Sea route, at roughly 15% of global refined-fuel supply. I’d treat that as a conditional total rather than a loss that’s already happened – the Gulf and Russian disruptions alone look closer to 11%, with the rest depending on whether the Red Sea threat becomes an actual closure rather than continued harassment. Either way, even higher crack spreads look likely, while crude prices will keep climbing.

The FT’s Christof Rühl has argued that the global economy has absorbed this shock far better than 1979 because of a production surplus prior to the war and efficiency. He claims high inventories built up before the war, oil stored on tankers, bypass pipelines, and non-Gulf production increases have kept the crude price subdued. Global production was likely in surplus through 2025, and much of that went to China building its strategic reserves. As explained previously, Chinese reserves are now being drawn down to cushion this crisis. Hence the surplus and the China buffer aren’t wholly separate, additive sources of resilience because to the extent that one became the other, counting both risks counting the same barrels twice. And it’s a buffer with a limit nobody outside Beijing can see.

Rühl calculates that oil intensity – barrels consumed per dollar of GDP – has reduced so much that prices would need to be roughly four times higher to do equivalent damage today. His numbers imply the gap still needing to be covered by lower demand or higher exports through the rest of the year is only 1.5 to 3.5 mb/d, which would be quite manageable. This is a similar argument to one made previously by Daniel Yergin, but I don’t find it any more convincing.

A problem with applying these so-called economy efficiency arguments to the current situation is that prices spiking and then partly recovering isn’t the same as the economy being able to withstand a supply-side shock. Companies used to keep spare stock sitting in warehouses and extra factory capacity sitting idle, and that spare capacity meant a supply disruption could be absorbed for a while before it actually affected output. Most of that spare capacity has been stripped out over the past few decades in the name of efficiency, especially through just-in-time inventory management. Rühl’s argument – that the economy needs fewer barrels of oil per dollar of output than it used to – is about efficiency when using oil, but it doesn’t reflect resilience, which is whether economies have sufficient buffer in reserves across multiple supply chains to absorb a shock.

The second problem is that his argument is about price, and part of what’s happening right now isn’t a price problem. Not to say prices are powerless as higher prices can redirect cargoes, push refiners elsewhere to raise runs, shift product yields, and draw down commercial stocks, which are the kinds of adjustments Rühl’s residual gap refers to. But none of that works instantly, and none of it works at all if a specific grade or a specific shipment is stuck behind a closed route with no substitute available in the necessary timeframe. Greater efficiency is not as useful if there are limited barrels or refined product to apply it to. If a factory can’t get the naphtha it needs because the ship carrying it is stuck in the Gulf, waiting for a shipping lane that may not open, no price increase or more rapid logistics network will fix that. A supply shock where the physical product simply isn’t available doesn’t respond to a higher price the way a normal shortage would. Price can only ration demand, not produce molecules. An aggregate barrels-per-GDP ratio doesn’t bring out that distinction, because it measures the economy’s average oil requirement, not whether any particular barrel can physically get to where it’s needed in time.

Behind his numbers is a third problem not acknowledged, and this is the increased demand for hydrocarbon products in modern economies. Global demand for oil has kept climbing to new records almost every year for decades, even as intensity improved, because the economy has grown faster than it’s become more efficient. But, most critically, the growth hasn’t been spread evenly across everything oil is used for. An increasing share of that growth is coming from petrochemical feedstocks rather than conventional transport fuels – the IEA now credits petrochemical feedstocks alone for more than half of the past year’s demand growth. That growth is in sources that didn’t really exist at scale four or five decades ago. Semiconductor manufacturing depends on specialty chemicals made from petrochemical feedstocks, and data centres are themselves built largely from petrochemical-based materials, compounded by the accelerating AI investment boom. All of that is subsumed within a ratio measured simply in barrels per dollar of GDP, which doesn’t reveal what the oil is being used for, how specialised those uses are, or how readily they can be substituted during a disruption. That demand growth resides particularly in petrochemical feedstocks such as naphtha, LPG and ethane, while the immediate shortage extends across diesel and jet fuel as well. These products have different transport routes, inventories and substitution constraints from crude, none of which is visible in an aggregate oil-intensity ratio.

History doesn’t back the smooth adjustment Rühl’s framework assumes, either. In 1973, a supply loss of only 4-7% of global output nearly quadrupled the price. In 1979, a smaller disruption still, roughly doubled it. Both were far larger price reactions than the size of the shortfall alone would predict, because oil demand is much less elastic than an efficiency ratio can capture on its own.

I am still monitoring three things that could yet prove me wrong on magnitude, and they each carry different probability and weight.

Hormuz reopening properly is the first, but it looks less likely by the week rather than more.

China is the second, with Chinese crude imports falling 41% year-on-year in June to a decade-low 7.12 mb/d, and roughly 28 mb drawn from stockpiles in that month alone. Reuters estimates China still added to reserves over the first half of the year as a whole, and puts total holdings at upwards of 1.2 billion barrels. It appears that China has been buying under $70 and pulling back above $80, so imports are therefore likely to recover in August and September reflecting purchases made during the temporary ceasefire when Brent fell towards $70. China’s reserve stays opaque on size and on where its own political floor sits, but it cannot draw down forever, and the timing of its permanent, full return is the largest single uncertainty hanging over price from here.

The third thing could potentially have a major effect on the thesis: the American strategic petroleum reserve floor. The SPR fell to 311.4 mb in the week to 17 July, the lowest since 1983. S&P Global stated the operational floor was only a few weeks away at the current drawdown rate, and as Japan and South Korea have already exhausted their own reserve releases from earlier in the year, oil traders are warning it wouldn’t take much to set off a jump to $200 given how thin global buffers have become. S&P Global was referencing a technical floor of 250 to 300 mb, consistent with both industry estimates and the original 1970s sizing studies. However, the US Energy Department told MarketWatch this week that the operational floor is as low as 70 mb. I’m not an engineer, and so what follows is me weighing competing claims rather than offering a technical judgement of my own.

Two things make me doubt the lower number, and neither is technical. It seems highly unlikely that successive US administrations would allow the market to believe for more than four decades that reserves are lower than they actually are. There is no motivation to maintain an illusion of smaller reserves, but every motivation to suggest higher reserves, either for market confidence or to project international strength and resilience. The Trump Administration, in particular, has made multiple optimistic claims with little foundational grounding, and never walks back the claims when they’re wrong. Tuesday’s Senate hearing was a perfect example. Senator Jon Ossoff asked Defense Secretary Hegseth three times whether his declaration, five months ago, that Iran’s military had been “destroyed” was accurate. Hegseth never said yes or no, pivoting each time to what the campaign had achieved rather than answering the question. A more pertinent example is the Energy Secretary himself, who oversees the SPR. He had been talking up a return to normal strait shipping only days before shipping collapsed again.

Of course, this is speculation and none of this proves the lower floor is wrong, yet there’s a further problem with the lower floor even if it turns out to be technically correct. Oil is withdrawn from the SPR by injecting water into the bottom of each cavern, which floats the oil up to the surface. This isn’t a straightforward tap that runs at a constant rate regardless of what’s left – the Government Accountability Office (GAO) reported in late May that the latest available assessment put the SPR’s effective drawdown rate at only around 61% of its original design capacity, and its effective fill rate at 56%, blaming a mix of construction outages, ageing infrastructure, deferred maintenance and brine-disposal limits rather than any single cause. Those estimates were based on conditions in December 2025, before the latest sustained drawdown, which GAO has warned would exacerbate existing limitations. So it’s not the case that the reserve is only approaching the design limitation in theory – it was already materially underperforming its specifications before the current drawdown, and may be underperforming them further now.

A reserve that releases oil slowly is of much less use in an emergency, the time when it is relied upon, so the SPR may stop functioning as an effective shock absorber before it stops functioning as a storage tank. And there’s a second problem with that. The whole idea of a strategic reserve is that you can draw it down fast in a crisis and then refill it once the crisis passes, ready for the next one. GAO puts the effective fill rate at only 0.44 mb/d, well short of what would be needed to rebuild the reserve quickly once a crisis passes, and replacing 200 mb at that rate would take well over a year of continuous operation even before accounting for congressional appropriation and procurement delays. GAO’s picture isn’t uniformly bleak, though, as most caverns are said to remain in reasonable condition, and the agency found some benefited from earlier drawdowns clearing salt blockages rather than being permanently damaged by them.

There remains a considerable amount of uncertainty here, some of it from incomplete information and some from a mix of overstatement and understatement depending on who’s speaking. The biggest news this week has been the lower SPR floor, but given who’s making the claim and when, I’d want it confirmed independently before treating 70 mb as the real number rather than the convenient one. In the meantime, the two shocks thesis remains live.

The MOU signed in mid-June set out a framework rather than a binding peace, since reopening a strait that had been mined and blockaded for months required more than just a signature. Following the MOU was a surge in traffic – 242 vessels transited the strait in the week to 28 June, against roughly 60 a week during the closure, and Iran exported around 70 million barrels during the truce. However, that surge was overwhelmingly the exit of a backlog already trapped inside the Gulf rather than a return of the two-way tanker process. Loaded vessels cleared cargoes that had been waiting for weeks and the empty tankers that needed to come back in to load the next round stayed deeply suppressed throughout.

The recovery never reached anything like normality, and its fragility was exposed early when an Iranian drone struck a cargo ship in the strait on 25 June, while the backlog was still clearing. The conflict that followed initially built steadily, if somewhat sporadically at first – further ship attacks on 6-7 July, and the US reimposing oil sanctions and striking more than eighty targets. In the past week or so it has gone well past where it stood before the MOU, with eight consecutive nights of American strikes on Iranian command centres, missile sites, coastal radar and the approaches to the strait. Iran has hit back across the region, targeting US facilities and American partners in Kuwait, Jordan, Bahrain, Qatar and Oman, while also claiming an attack on a US base in Saudi Arabia.

Tehran has now formally suspended its participation in the MOU altogether, with Supreme Leader Mojtaba Khamenei declaring that the repeated breaches have shown Trump’s signature to be “utterly worthless and devoid of credibility”. Iran puts its own toll at 50 dead and more than 500 wounded since fighting resumed. Two US service members were killed by an Iranian strike on American forces in Jordan on the night of 17-18 July, bringing the total to sixteen US deaths over the course of the war.

The blockade has moved from threat back to enforcement over the same stretch. American forces disabled the oil tanker Belma with Hellfire missiles on 15 July as it tried to run the blockade toward Kharg Island, the terminal which handles roughly 90% of Iran’s crude exports. This was the first time this round of the blockade had been enforced with weapons, rather than through warnings, boardings, and orders to turn back. Tanker traffic through the strait has subsequently fallen to its lowest level since May – only six oil and gas tankers crossed on 12 July, and only three commodity tankers on 16 July. The US Energy Secretary, however, claimed on Friday that American assistance was still helping roughly 7 mb/d leave the Gulf, which is a much broader and less verifiable figure than the confirmed transit counts, and one increasingly complicated by vessels operating without visible tracking signals.

A toll proposal came and went almost as quickly as it appeared. Trump floated a 20% fee on strait cargo on 13 July and dropped it the next day for vague talk of investment deals with Gulf states, the third toll idea aimed at this waterway in four months, after Iran’s own transit fee proposition and Trump’s floated joint venture with Iran in April. None of them has lasted more than a few days.

Trump has warned of striking Iranian power plants and bridges repeatedly since January, usually as a deadline that passed without full follow-through. This time it was carried out: on 17 July, US strikes hit six bridges in Hormozgan province, killing at least seven people and wounding twenty, on the roads connecting to Bandar Abbas – the IRGC naval base city, and the key target for any US ground operations, as argued in April, as opposed to Kharg Island itself. A maritime control tower at Chabahar, further along the Gulf of Oman coast, was destroyed the same night. Kharg seizure talk has also returned, although now with Trump suggesting the ground campaign might fall to “other people” rather than American troops. Reporting has rightly been pushing back, noting that holding the island would likely require a far larger force than currently assembled, regardless of who provides the troops.

The crude price continues to be the part of my forecast that has not yet matched my calculations. Yet, Brent has now fully round-tripped the entire post-MOU relief rally – from $76 on 10 July to $88 by the 17th, up around 16% for the week – and the market mechanics underpinning the increase look closer to the ones I have previously described rather than just a passing scare. Prompt Brent flipped from contango into nearly $2.50 a barrel of backwardation mid-week, a clear signal of immediate physical tightness rather than an insurance cost specifically. Refined product cracks have moved further still, with diesel margins above $85 a barrel and gasoline margins around $55, both back toward the extremes reached earlier in the crisis. Barclays’ research desk described much the same picture in less colourful language this week: SPR releases have shielded the crude market, but commercial inventories “do not look comfortable”, and the bank is holding its $96 full-year forecast for 2026 regardless of where spot sits today. The SPR is down almost 99 million barrels since the war began, to 316.5 million, and Cushing stocks remain close to the operational floor tested earlier in the year.

What has kept price below the earlier calculations is not that the assessment was wrong, but that the offsets have proven larger or slower to unwind than modelled. Chinese imports have been running well below their five-year average since April and fell to around 40% of pre-war levels in June, but whether and how far they recover remains an ongoing uncertainty – so one of the largest cushions still in place is one whose timing nobody can currently predict. At the same time, Iraq has been building additional bypass routes, trucking fuel oil primarily through Syria to Baniyas and smaller volumes through Jordan to Aqaba – workarounds rather than durable pipeline capacity, though the Baniyas trade alone was large enough to make Syria the Middle East’s largest exporter of high-sulphur fuel oil in June. More permanent pipeline expansion is being accelerated across the Gulf states generally, though most of that capacity is still years from having a significant impact.

My assessment and forecasts have framed the whole period as a race between reserves and restoration – whether the strait would return to normal before the temporary buffers ran out. That race is not yet over, given Chinese and US reserves still hold, but it looks increasingly likely Hormuz will lose. Transit hasn’t normalised – it has collapsed again and worse in places than before the MOU. What steps in as a substitute to chase the reserve runout early lead is how fast bypass routes – Iraq’s trucking workarounds and the wider Gulf pipeline build-out – can scale to cover what the strait no longer reliably carries. Whether the forecast turns out to be early rather than wrong depends on which of those two finishes first, but on current form it is difficult to see the new entrant close the gap in time.

The previous article, PE1, established Proxy Australia with a circulating money stock of $3.0 trillion and GDP of $2.0 trillion, giving a velocity of 0.67. This article explains what that last figure means, why it is neither stable nor easily controlled by policy, and why that instability effectively demolished one of the most influential economic theories of the twentieth century.

The identity

Velocity emerges from a relationship known as the equation of exchange, written as M × V = P × Y. M is the money stock, V is how many times each dollar turns over in transactions in a given period, P is the price level, and Y is real output, so P × Y is nominal GDP. The equation says that total spending in the economy (money multiplied by how often it circulates) must equal total output (prices multiplied by quantities). In Proxy Australia: $3.0 trillion multiplied by 0.67 gives $2.0 trillion of nominal GDP.

This identity is always true, in the same way that every sale has a buyer and a seller. It is a definition, a way of expressing what must be the case once we have decided what we mean by money, spending, prices, and output. The interesting questions begin when we ask what determines each of the four variables and how they interact.

The intuition behind velocity is easier to grasp with a simpler example. Imagine a village with ten dollars and ten transactions a year, each worth one dollar: GDP is ten dollars, money is ten dollars, velocity is 1. Now suppose those same ten dollars circulate twice as fast – the money completes twenty transactions over the year. GDP doubles to twenty dollars without any new money being created. Velocity has done the work that more money could have done instead.

This substitutability – velocity and the money supply as alternative routes to the same nominal GDP – is the central insight discussed here, and the one that has generated the most controversy.

Friedman and the velocity assumption

Milton Friedman spent much of his career building an ambitious intellectual structure on top of that simple identity, and the stability of velocity was its foundation. His argument, developed across decades but stated most influentially in his 1963 work with Anna Schwartz, A Monetary History of the United States, was that velocity was not constant but was sufficiently stable and predictable to serve as the basis for monetary policy.

If V moves slowly and predictably, then controlling M controls nominal GDP.

The central bank, on this view, holds the lever that determines nominal economic outcomes.

From this he drew a policy prescription: central banks should target a fixed, modest rate of money supply growth – something like 3% to 5% a year – rather than trying to fine-tune the economy through discretionary intervention. This was monetarism, and through the 1970s it was enormously influential, in part because it offered a clear explanation for the inflation of that decade. Too much money had been created, particularly after the US abandoned gold convertibility in 1971, and the consequence was higher prices. The argument was coherent, the historical evidence that Friedman and Schwartz assembled was impressive, and the policy prescription was clear.

As a long-run statement, it is essentially correct. Sustained high inflation requires sustained money supply growth – hyperinflations always involve dramatic monetary expansion, and price stability over long periods is associated with controlled money growth. The identity demands it: if nominal GDP rises persistently while velocity is roughly stable, the money supply must be rising. Friedman was right about this.

Where he was wrong, or at least where the monetarist programme ran into serious trouble, was in assuming that velocity was stable enough to make money supply targeting a reliable instrument. The 1980s tested that assumption and it failed.

The 1980s: when velocity moved

Margaret Thatcher adopted monetary targets in Britain in 1979 and Ronald Reagan’s Federal Reserve chairman Paul Volcker pursued them in the United States from 1980. Both were acting on broadly monetarist principles: control the money supply and inflation will follow. Both did break inflation, but not quite through the monetarist mechanism. The money supply numbers were erratic and often missed their targets. Rather, what actually drove disinflation was very high interest rates that came with tight monetary policy, which directly crushed demand.

More damaging for the theory, velocity behaved badly. As financial deregulation proceeded through the 1980s, new monetary instruments proliferated, such as money market funds, interest-bearing current accounts, and novel credit products. The relationship between any particular measure of M and nominal spending became unstable. People shifted money between instruments in ways that scrambled the data. Velocity of narrow money measures rose while velocity of broad measures fell, and neither moved as the models predicted.

By the mid-1980s, monetary aggregate targeting had been dropped without ceremony in both countries, replaced by interest rate targeting. The central bank would set the price of borrowing rather than the quantity of money, and let the money supply be whatever the economy demanded at that price. This is still how central banks operate today, though they rarely say so in those terms. Monetarism as a policy programme was brought undone by financial innovation.

Friedman himself was not oblivious to this. He acknowledged that the stable money demand function he had estimated in the 1950s and 1960s became unstable in the 1980s, though he attributed this partly to the distortions of the high-inflation period rather than rejecting the framework entirely. His supporters argued the targets had been abandoned too soon or implemented imperfectly. These debates continue among monetary economists, yet the practical result was clear: the central banks of the world stopped targeting the money supply.

What the popular reading of Friedman misses

The version of Friedman’s insight that circulates in public debate – “printing money causes inflation” – strips out the two most important qualifications in the original argument.

The first is velocity. If the money supply doubles but velocity halves, nominal GDP is unchanged and there is no inflation. And that is not a theoretical curiosity as it happened after 2008. The US Federal Reserve, the Bank of England, the European Central Bank, and the Bank of Japan all conducted large-scale asset purchases – quantitative easing (QE) – that dramatically expanded the deposits banks held at their central banks. Whether and how far that expansion fed through into broader money circulating in the economy depended on whether banks lent and whether borrowers borrowed, and for much of the post-crisis period they did neither aggressively. On a crude reading of “more money equals more inflation”, price levels should have surged. They did not, and for a decade central banks struggled to get inflation up to their 2% targets rather than down from an overshoot.

The reason was lower velocity. Banks sat defensively on the new money rather than lending the funds out aggressively, and households and businesses – recovering from the shock of the financial crisis – paid down debt rather than taking on new borrowing. Money turned over slowly, hence the new money supply from QE – central bank asset purchases – translated into far less new spending than the headline numbers implied.

The second qualification is spare capacity – whether the economy can respond to increased spending with more output rather than higher prices. Friedman’s framework acknowledged that in the short run, if there were unemployed workers and idle factories then more money could raise real output Y rather than prices P. The purely monetary inflation story applies only when the economy is at or near capacity, in which case the additional spending cannot call forth more goods or services and must instead bid up their prices. In Proxy Australia’s terms: if 14 million workers are employed and output is at its maximum, any increase in M × V shows up as higher prices. If there are unemployed workers, some of it shows up as higher output instead. The distinction between those two cases is one the simple “printing money causes inflation” formula erases entirely.

Inflation, the binding real constraint on a monetary sovereign’s spending, will get a full article of its own – PE13 – once the model has a government sector and banking system in place. What is important here, though, is that money supply alone does not determine prices, because velocity affects the relationship between money and spending, and real capacity affects that between spending and prices.

The 2021 inflation: a test case

The inflation in 2021 and 2022 prompted a revival of monetarist arguments, with some justification. Governments had spent heavily during the pandemic, central banks had kept interest rates near zero and continued asset purchases into the recovery, and so money supply growth in the major economies ran at rates not seen since the 1970s. Some economists argued the resulting inflation was exactly what monetarism predicted.

The argument is not wrong, but it is incomplete. The inflation of 2021-22 had clear supply-side components – energy prices driven up by the Russo-Ukraine war, supply chains disrupted by the pandemic’s uneven unwinding, and labour markets that tightened faster than expected as activity recovered. Yet these were not monetary phenomena in the Friedmanite sense; rather they were supply shocks, constraints on the supply of real goods that pushed prices up regardless of what happened to money. In the equation of exchange, they are a constraint on Y – real output cannot expand quickly enough to meet demand – which forces P upward even if M and V are held constant.

What the 2021 episode illustrated was the interaction of monetary and supply-side factors rather than the dominance of either. The monetary stimulus of 2020-21 raised nominal demand, but supply constraints prevented output from expanding to meet it, and so prices rose. A purely monetary account explains the demand side but needs supplementing to account for the timing and sectoral pattern of price increases. A purely supply-side account explains why prices rose in specific areas but struggles to explain the breadth of the inflation. The identity M × V = P × Y correctly shows both sides are important in diagnosing the outcome.

Why velocity moves

Velocity is not a fixed number. It varies with confidence, interest rates, the composition of the money stock, and institutional behaviour.

When households and businesses are confident, they spend readily, hence money changes hands quickly and velocity is high. When they are anxious, they accumulate money balances rather than deploying them and velocity falls. Keynes called the demand to hold money rather than spend it “liquidity preference”, and a rise in liquidity preference is arithmetically equivalent to a fall in velocity.

Interest rates affect velocity through the opportunity cost of holding cash. When rates are low, holding money is cheap – there is little return foregone by sitting on a cash balance – and velocity tends to be lower. When rates are high, the cost of holding idle cash is greater, which pushes money into active use so velocity tends to be higher.

The composition of the money stock is important too, which is why the choice of monetary aggregate changes the velocity figure. A narrow money measure like M1 – cash and current accounts – has a higher velocity because current account balances are actively used in transactions. A broad measure includes term deposits and other instruments that turn over slowly, hence the velocity figure is lower. Proxy Australia’s circulating money stock of $3.0 trillion and velocity of 0.67 reflects the weight of less-active holdings in the aggregate, which is why the money stock is larger than annual GDP rather than smaller.

What velocity cannot tell us

Velocity is an accounting residual, not a causal variable. M × V = P × Y is always true by definition – it expresses a relationship that must hold once we have defined our terms – but it does not specify the direction of causation. A rise in nominal GDP could pull velocity up if the money stock is slow to respond; an increase in spending confidence could push velocity up and nominal GDP with it; or both could move together driven by some third factor. The identity describes the relationship without explaining it.

This limits what policy can do with velocity as a target. Central banks can influence the money supply M, and can nudge interest rates which affect velocity indirectly, but they cannot set velocity directly. Velocity is an emergent property of billions of spending decisions, not a dial that policy turns. This is the lesson the 1980s taught monetarism.

Velocity in Proxy Australia going forward

In the model, velocity is set at 0.67 and held constant throughout the series. This is a simplifying assumption: as discussed above, in reality, velocity varies with confidence, interest rates, and the composition of the money stock. Holding it fixed allows isolation of the effects of other changes and tracing them cleanly, which is the purpose of the model. The dynamics that move velocity in practice – credit expansion, financial stress, shifts in liquidity preference – are covered in later articles through the mechanisms that actually produce them, rather than through velocity as a variable in its own right.

Parameters carried forward

Population: 25 million

GDP: $2.0 trillion

GDP per capita: $80,000

Circulating money stock: $3.0 trillion

Velocity: 0.67

Employed: 14 million; Students: 4 million; Retired: 5 million; Unpaid caring and household work: 2 million

Every model is a simplification. The question is whether the simplifications are appropriate, serving the argument or obscuring it. This article explains what Proxy Australia simplifies and why, what the model can and cannot do, and establishes the parameters that carry forward through the series unless explicitly revised.

Why a stylised economy rather than real data?

There are two tempting alternatives to the approach taken here. The first is pure abstraction – a generic economy with round-number parameters chosen for arithmetic convenience, untethered from anything a reader might recognise. The second is empirical precision – every parameter sourced from an official dataset, complete with decimal places and a vintage date.

Both are worse for the purpose at hand.

Pure abstraction strips out real-world intuition. When a model sets the workforce at 50% of the population because that ratio is convenient, readers lose their footing – they cannot sense-check outputs against anything they know about the actual world, and the model ends up teaching its own internal logic rather than economics that transfers to reality.

Empirical precision creates the opposite problem: false authority. A model reporting GDP as $2.248 trillion – the ABS figure for a specific quarter – implies it is a faithful representation of Australia in that quarter, yet it is not. GDP revisions are material, the workforce shifts over years, and monetary conditions change quarter to quarter. A model pinned to a particular vintage of data will be “wrong” in a distracting way the moment that data is revised, drawing objections about the numbers rather than engagement with the economics.

The stylised approach takes a middle path. Parameters are chosen to be plausible rather than precise – close enough to Australian reality that results feel intuitive and can be compared to lived experience, yet rounded sufficiently that no one should mistake them for official statistics. The calculations also stay comprehensible: $2.0 trillion in GDP and $3.0 trillion in money stock are numbers a reader can hold in their head and check against without reaching for a calculator. The goal is a model that aids understanding, not one that forecasts.

What “roughly grounded in reality” means

Each parameter falls within the range of plausible Australian values over recent years. “Roughly grounded” means a reader familiar with Australian macroeconomic data should find nothing surprising or implausible here. It does not mean the parameters are calibrated to a specific year, that they will match ABS or RBA publications, or that deviations from published figures are errors. They are deliberate choices about modelling convenience that preserve economic realism without claiming statistical precision.

What this model can and cannot do

The model is designed to illuminate the mechanisms of monetary economics: how money is created and extinguished, how the quantity of money in circulation relates to nominal output, and how changes in the money stock propagate through an economy. Because the parameters are grounded in a recognisable economy, results expressed in terms of Proxy Australia are legible to anyone with a passing familiarity with Australian economic commentary.

What it cannot do is predict. There is no time dimension, no stochastic elements, no mechanism for incorporating expectations or confidence effects. It will not tell you what the RBA should do at its next board meeting.

The initial setup also omits a government sector and an external sector – no fiscal policy, no public debt, no exports or imports, no exchange rate. Both are features of the starting model rather than permanent omissions, and each will be introduced in later instalments as the model is progressively built out.

Finally, the model works with aggregates. It will tell you something about the total money stock or the aggregate price level; it will not tell you who holds the money, who benefits from inflation, or how monetary conditions affect different income groups differently.

A model that is clear about its limits is more persuasive than one that oversells itself. These boundaries define the scope within which the model’s outputs can be trusted.

The parameters

The baseline economy has a population of 25 million and GDP of $2.0 trillion, implying GDP per capita of $80,000. That per capita figure is broadly consistent with Australian National Accounts over recent years and gives results that feel intuitively plausible without being tied to any single quarter’s data.

The 25 million population breaks into four groups: 14 million employed, 4 million students, 5 million retired, and 2 million in unpaid caring and household work. The employed workforce represents a participation rate of 56% of total population, a figure that makes sense once the school-age population, the student cohort, and a growing retiree base reflecting demographic ageing are set aside.

The monetary parameters follow from the equation of exchange: M × V = P × Y. Here M is the circulating money stock, the total quantity of money moving through the economy at this stage of the model, before bank credit creation has been introduced; V is velocity, meaning how many times, on average, each dollar turns over in transactions in a given period; and P × Y is nominal GDP, the price level multiplied by real output, which together give total output valued at current prices.

With nominal GDP at $2.0 trillion and a circulating money stock of $3.0 trillion, velocity is 0.67. This is the right order of magnitude for a broad monetary aggregate in a developed economy – a large share of the money stock sits in savings instruments and less active holdings that don’t turn over quickly, which is why the aggregate figure is larger than annual GDP rather than smaller. The number is rounded for modelling convenience rather than calibrated to a specific RBA data series.

The baseline economy begins from a position where no forces are pushing it away from its initial state:

This is a deliberately simple starting point, not an idealised policy target and not a description of Australia at any particular moment, but a neutral initial state from which changes can be introduced, traced through the model, and understood.

Each subsequent instalment adds one layer to this foundation. The next article, PE2, covers the equation of exchange – what velocity means, why it matters, and why it is neither fixed nor easily controlled by policy. That needs to be in place before the government and banking articles complicate the picture further. The parameters established here carry forward unless an article explicitly revises them, and when they are revised the change will be stated clearly so the model remains internally consistent throughout.

Why this series exists

Most economic commentary starts from conclusions. The deficit is too high. Interest rates need to rise. The government is spending beyond its means. These conclusions arrive already formed, drawn from textbooks, historical analogies, and political instinct – and when the world doesn’t behave the way the conclusion predicted, the response is usually to add a qualification rather than to question the underlying reasoning.

This series starts somewhere different: from the mechanics. Not from a school of thought, not from a political position, and not from the assumption that the standard account is correct. Instead, from the focus on how a monetary economy actually works, built up piece by piece, with the numbers set out clearly and the accounting checked at each step.

That approach is important for reasons that go well beyond academic tidiness. Consider the claims that circulate in almost every budget debate: the government must live within its means; taxes fund public spending, so if we cut taxes we can’t afford the services; deficit spending is inflationary; the national debt is a burden we’re passing to our children; we simply can’t afford it. These are not fringe positions. Rather, they are the default language of fiscal policy across most democracies, repeated by finance ministers, central bankers, and newspaper editors as though they were axiomatic rather than conclusions that depend on how the monetary system actually works.

When the mechanisms behind them are examined carefully, several turn out to be wrong in critical ways – not necessarily wrong in direction, but wrong about the causal chain, which means the policy prescription that follows can be backwards. Austerity imposed on a government that issues its own currency is not the same thing as a household cutting its credit card bill, however similar the language sounds. Tax revenue does not fund a sovereign government’s spending in the way a salary funds a household’s. Deficit spending is not inherently inflationary – whether it is depends on where the spending is directed and on where the economy is relative to its real productive capacity. Getting the mechanisms correct determines whether the prescription matches the diagnosis. It’s not merely a technicality, as the consequences of that mismatch affect real people.

Most interesting, the underlying premise of the series extends into broader issues. Monetary economics is not merely a domestic matter. How currencies relate to each other, which countries can finance themselves freely and which cannot, how sanctions weaponise payment infrastructure, why the Eurozone’s design produced the crises it did, what reserve currency status actually confers – these are questions of international political economy and geopolitics as much as they are questions of economics. The same analytical failure that produces bad domestic fiscal policy produces bad analysis of sovereign risk, currency crises, and great-power competition over monetary order.

The US blockade of Iranian ports is a case in point. The strategic theory behind the blockade was that cutting Iran’s export revenues would deprive its government of the funds it needed to operate. However, a government that issues its own currency does not fund itself through export earnings. Exports are only necessary to pay for imports, so a blockade restricts access to real goods, technology, and components that cannot be obtained domestically. That is a constraint, but a different one from the fiscal constraint the policy assumed, and a blockade strategy designed around such a basic misunderstanding was always going to underperform.

I should say up front that this series is not only an attempt to teach. It is also a way of thinking things through. Working step by step through a simplified model forces a precision that intuition and pattern-matching do not. When the numbers don’t add up, or an argument that seemed coherent turns out to be inconsistent with an accounting identity, the model reveals it. This is more useful than discovering the error after the conclusion has already been formed. I have found, working through these articles, that the process of building the model has changed some of my own views, confirmed others, and in several cases revealed that what I thought I understood, I had understood only approximately. And that is the point. The aim is not to transmit a settled set of correct views about economics, but to demonstrate a way of reasoning about it that works when the next crisis arrives and the pattern-matching fails.

What the model is

A simplified model of a fictional economy – Proxy Australia – is implemented by the series. It is built in layers, starting from first principles and adding one mechanism at a time. The economy is based roughly on Australian numbers, grounded in reality but simplified, as precision would just add complexity without adding clarity. The model is not Australia precisely, but it is close enough that the institutional features are recognisable and the numbers are defensible. The series is built on a layered structure. The banking system cannot be understood properly without the government sector in place first. Unemployment cannot be analysed fully without the banking system already functioning. The open economy only makes sense once the closed economy is solid.

The series will also comment where it agrees and disagrees with existing schools of thought. Modern Monetary Theory – MMT – gets significant things right about how monetary sovereign economies work, but it also overstates its case or explains itself through analogies that mislead as often as they illuminate. Orthodox economics similarly gets important things right, but gets other things wrong that have caused measurable damage, particularly the persistent application of household-budget logic to monetary sovereign governments. The aim throughout is accuracy about the mechanisms, not advocacy for any school.

A note on monetary sovereignty

Before the series begins, one foundational condition needs to be stated. Every article in this series describes a monetary sovereign economy.

A government is a monetary sovereign if it issues the currency it borrows in. A Quick Note on Monetary Sovereignty explains this condition in full. The short version is that Proxy Australia issues its own currency, that currency floats freely, and it does not borrow in any other currency. A government borrowing in the currency that it issues itself faces no financial constraint analogous to a foreign-currency borrower – it cannot be rendered insolvent by an inability to repay or a refusal of external debt holders to refinance. Self-imposed constraints such as debt ceilings or appropriations rules can interrupt payments, but those are political choices rather than external financial limits. The binding constraint for currency issuers is inflation, which reflects the limits of real productive capacity. Countries in this category are the US, UK, Australia, Japan, Canada, and others.

Currency users face different constraints from currency issuers, and are closer to the household analogy. Eurozone member states – France, Germany, Italy, Spain, and the rest – share the euro but cannot create it. Countries that have dollarised or operate a currency board are in a similar position. These economies have different mechanics, with actual financing constraints rather than just political ones. Bond markets can impose solvency-like pressure and the household-budget analogy is less a metaphor than a reasonably accurate description of their position. The series will address currency users in a separate extension series. For now, every article in the foundational series rests on the monetary sovereignty condition.

Useful background to help thinking through this distinction, before any of the model-building begins, is The Banker Never Goes Broke. This is a standalone piece using Monopoly to illustrate why the currency issuer and the currency user face fundamentally different constraints. The Monopoly banker cannot run out of money, because the banker creates the money the game runs on. Players can go bankrupt, but the banker cannot. That distinction – issuer versus user – is the conceptual spine of everything that follows.

The foundational series: PE1–PE17

Layer 1: The real economy

PE1 – Assumptions and Parameters

Every model is a simplification. This article will explain what Proxy Australia simplifies and why, what the model can do and cannot, and establishes the parameters that remain fixed throughout the series unless explicitly revised. These parameters include population 25 million; GDP $2.0 trillion; circulating money stock approximately $3.0 trillion, giving a velocity of 0.67; and the workforce split into approximately 14 million employed, 4 million students, 5 million retired, and 2 million in unpaid caring and household work. The economy starts as a closed economy at full employment. There is a stock of money in circulation – how it came to exist is not important for the series. The government sector and banking sector will both be introduced in later articles, at which point the composition and creation of money will become central to the analysis.

PE2 – Money and Velocity (MV = PY)

What money does in an economy depends not only on how much of it exists but on how quickly it circulates. This article will explain the equation of exchange – the relationship between the money stock, velocity, and nominal GDP – and why velocity is neither fixed nor easily controlled by policy. That needs to be covered before the government and banking articles which complicate the picture further.

PE3 – What Money Is

An exploration on the nature of money itself: why it is a liability of whoever issues it, why its acceptance depends on the authority that stands behind it and the obligations it extinguishes, and why the distinction between money-as-commodity and money-as-credit is not academic but structural. The article will also address one of the most persistent myths in economic textbooks – that money evolved from barter as a more efficient medium of exchange. This will lay the conceptual groundwork for the taxation and bond issuance articles that follow.

Layer 2: The government sector

PE4 – Government Spending

The government will begin spending, but without taxation yet. This article will ask where the money comes from, follow it through the model’s accounts, and show what happens to the money stock and the private sector’s financial position when the government spends before collecting any taxes. This is where the spend-then-tax sequence – one of MMT’s genuine insights, first articulated clearly by Warren Mosler – is introduced and explained.

PE5 – Taxation

Taxation will be introduced. The article will draw on Mosler’s business-card parable from his Soft Currency Economics text to explain why taxes do not fund government spending. Rather, they create the demand for the currency the government has already spent into existence. The scope condition in the series is crucial here. The spend-then-tax sequence holds for currency issuers, but for currency users, spending first requires acquiring the currency from elsewhere – through taxation, borrowing, or transfers – since they cannot create it themselves.

PE6 – The Government Deficit

With spending and taxation both in the model, Proxy Australia will next run a deficit: $120 billion in spending against $100 billion in tax. This article will examine what a deficit is, what it is not, and why the household-budget analogy fails as soon as accounting concepts are applied. The government deficit is, by accounting identity, the private sector’s surplus – in a closed economy, every dollar the government spends in excess of what it taxes is a dollar that has remained in private hands.

PE7 – Bond Issuance

The government will next issue bonds. This article will explain how bond issuance converts money in the economy into interest-bearing securities, and why this is an asset swap rather than borrowing in the household sense. Bond vigilantes, debt sustainability, and the crowding-out argument will all appear here. The loanable-funds version of financial crowding out occurs where government borrowing competes for a fixed pool of savings and drives up rates, but it rests on assumptions that do not hold for a monetary sovereign. Real-resource crowding out at full employment is a legitimate concern, but it is different and will be addressed separately.

Layer 3: The banking sector

PE8 – Banking and Credit Creation

Banks create deposits when they make loans. Banks do not create money by lending out deposits – that gets the causality backwards. This article will explain the mechanics of bank credit creation and address the loanable funds model – the standard textbook picture of banks as intermediaries between savers and borrowers – which does not capture how modern banks actually work. The article will also establish the distinction between government-created money – net financial assets for the private sector – and bank-created money, which is matched by an equal private debt obligation that must eventually be repaid.

PE9 – The RBA and Reserve Requirements

The central bank does not control the money supply. Through the cash rate it sets the price of overnight settlement balances. Banks hold these balances with the central bank because of the domestic payments system. The central bank supplies whatever quantity of funds are required to clear the balances at its chosen interest rate. This article will explain the mechanics of how this works and why, what the RBA (Reserve Bank of Australia) actually does and what monetary policy can and cannot achieve through the interest rate channel.

PE10 – Private Debt Dynamics

Unlike government spending, bank loans must be repaid, and repayment destroys the money the loan created. This article will introduce Hyman Minsky’s financial instability hypothesis. Minsky observed that periods of stability are self-undermining as calm conditions encourage progressively riskier borrowing, which plants the seeds of the next crisis. Professor Steve Keen extended this insight, arguing that it is the change in private debt, rather than its level, that drives aggregate demand at the margin. This will be the theoretical foundation for the financial crisis article later in the series.

Layer 4: Scenarios

PE11 – Unemployment: Cyclical

Full employment will now end. Consumer confidence falls, spending contracts, and workers lose jobs through no fault of their own and no change in the economy’s productive capacity. This article will examine cyclical unemployment through the output gap that opens, the automatic fiscal stabilisation that follows, and the limits of monetary policy as a recovery tool when the problem is a demand shortfall.

PE12 – Unemployment: Sectoral

Not all unemployment is cyclical. This article will model a concentrated shock in one sector of Proxy Australia’s economy, with the remaining sectors at or near capacity. The policy implications differ materially from the cyclical case – aggregate demand stimulus is a blunt instrument when the problem is mismatch rather than deficiency – and the article will examine both what fiscal and monetary policy can do and what they cannot.

PE13 – Inflation

Inflation is the binding real constraint on a monetary sovereign’s spending, resulting from the availability of real resources, not from the availability of money. This article will address Milton Friedman’s claim that inflation is “always and everywhere a monetary phenomenon”, and argue instead that whether spending is inflationary depends on where the economy is relative to its productive capacity. The Monopoly analogy will show why running out of real resources, not running out of money, is what actually creates inflation pressure.

PE14 – Financial Crisis

A Minsky moment will arrive in Proxy Australia. Private credit has expanded through the speculative and Ponzi phases, and asset prices have been sustained by refinancing rather than income. When the refinancing chain breaks, the contraction will be sharp, non-linear, and not amenable to monetary policy alone. This article will run the crisis template through Proxy Australia, showing how the Minsky dynamic from PE10 resolves into crisis and what role the government and central bank play in containing it.

Layer 5: The open economy

PE15 – Trade

The external sector will now appear. Proxy Australia will engage in imports and exports. The article will follow the flows: what a current account deficit means for domestic financial claims and foreign ownership of assets, what exports actually represent for a monetary sovereign, and where the bilateral trade deficit numbers that dominate political debate fit, and do not fit, into any coherent analysis of whether trade is beneficial or harmful.

PE16 – Foreign Direct Investment

Next will be consideration of foreign sources of capital. This article will explain the difference between foreign direct investment (FDI) and portfolio investment. In particular it will clarify what FDI provides and what it doesn’t. Capital is not something a monetary sovereign needs – foreign capital arriving to build a factory gets converted into the local currency and spent in the local economy. What FDI actually brings that’s worth having is the technology, skills, and knowledge transfer that comes attached to it, and that needs to be distinguished from the headline figure itself.

PE17 – Monetary Sovereignty in Practice

The capstone of the foundational series. Monetary sovereignty turns out to be not a binary condition but a spectrum with three layers: the federal government’s full operational sovereignty, the banking system’s foreign-currency wholesale funding vulnerability, and the exchange rate and market pressures that constrain fiscal behaviour in practice without creating an operational financing constraint. Monetary sovereignty removes one specific limit – the financing constraint – while leaving real resource constraints, external vulnerabilities, and political constraints fully intact.

Cross-references to existing articles

Several articles published before this series began cover themes the series addresses in more depth. The most directly relevant:

The Banker Never Goes Broke – the Monopoly primer on the issuer/user distinction, the natural starting point before PE1.

A Quick Note on Monetary Sovereignty – the foundational scope condition explained in full.

The Wrong Variable – bond markets don’t assess sovereign credit worthiness – they assess supply, and as a flow variable this is where the fiscal deficit becomes important.

The Prescription Doesn’t Match the Diagnosis – orthodox fiscal prescriptions misapplied to monetary sovereigns.

Exports Are Only Necessary to Pay for Imports – why exports aren’t revenue, and why that affected the US blockade of Iran.

Why Debt/GDP Is Witchcraft – the standard debt sustainability metric and its problems.

The Economy Is Not Constrained by Money – the resource constraint as the binding limit.

Extensions to the foundational series

The following articles are possible extensions of the foundational series. They are listed here so readers can see where the series is likely to head, but nothing definitive has been drafted yet.

The Loanable Funds Model and Why It Fails

Fractional Reserve Banking – the Myth that Doesn’t Die

Debt Sustainability

The Natural Rate of Interest

NAIRU and the Natural Rate of Unemployment

Quantitative Easing

Gold Standard versus Fiat Money

Extension series: currency users

The foundational series describes monetary sovereigns. A parallel extension series will examine currency users – economies where there is a financing constraint, the central bank’s role is different, and the household-budget analogy is a reasonably accurate description rather than a misleading one.

This will examine the Eurozone in detail: a monetary union without fiscal or political union, the ECB as a currency issuer whose support for member-state debt is contingent rather than automatic, and the 2010–2012 sovereign debt crisis as the proof of concept. This series will also cover dollarisers and currency boards, developing-country foreign-currency debt, and the spectrum of monetary sovereignty between the full-sovereign and the full-user cases.

International political economy

After the foundational series and some of its extensions, the project will move to international political economy, the layer at which monetary arrangements meet state power, geopolitical competition, and the distribution of advantage among nations.

The international monetary order is not a neutral backdrop against which geopolitics plays out. It is itself an instrument of state power, and its design at any point in history reflects the distribution of power among states as much as any rational economic calculation. Major monetary regime changes have almost always followed wars or the near-failure of the prior arrangement under the stress of conflict: the gold standard did not survive the First World War’s financing demands intact; Bretton Woods was designed in the ruins of the interwar system that war debts and reparations had destroyed; the dollar’s reserve status was built substantially on the US’s position as the only major economy left with productive capacity and gold reserves after 1945.

That history is directly relevant to the present. The freezing of Russian central bank reserves after 2022, the acceleration of reserve diversification away from dollar and euro assets, China’s gradual renminbi internationalisation, the use of SWIFT exclusion as a policy tool, the return of tariff policy framed in explicitly mercantilist terms – these are simultaneous live tests of the international monetary order’s durability, and they are understandable only through the kind of analytical framework the foundational series is designed to build.

Topics for the IPE series also may include: reserve currency status and what it actually confers on the issuer; the Eurozone’s political economy and why member states accepted monetary union without fiscal union; developing-country debt crises and the structural vulnerability of peripheral currencies; sanctions and the weaponisation of payment infrastructure; China’s currency management and reserve accumulation strategy; and the impossible trinity of fixed exchange rate, free capital movement, and independent monetary policy.

The history sub-series

An examination of how monetary economies have actually worked across history. The purpose will be diagnostic, rather than narrative, to identify a specific and recurring analytical failure, in which a fragment of historical knowledge, detached from its mechanism, is misapplied as a general law.

Each article will classify its episode under one of three headings.

Genuine regime change: the system really did work differently, and changed at an identifiable point for identifiable reasons.

Misremembered mechanism: the system worked differently, but not in the way commonly described, so the lesson drawn from it is wrong even though the underlying change was real.

Retrospective fiction: the system never worked the way it is now assumed to have worked, even at the time.

Articles planned for the history sub-series include:

Taxation in pre-modern and ancient economies, including Rome – why specie as the monetary base changed the fiscal logic fundamentally, and how tax farming differed institutionally from a modern tax authority

Mercantilism and the origins of trade confusion – the gold-accumulation logic that made “exports good, imports bad” coherent in its original context, and the residue it left in trade policy thinking long after that context vanished

The Bank of England’s founding and early debt financing – war finance, permanent national debt, and the gradual working-out of what sovereign debt in one’s own currency actually entails

The classical gold standard (c. 1870-1914) – what it actually constrained, how adjustment was supposed to work, and why it held together better for some countries than others

The interwar restoration and its failure – the attempt to return to gold after 1918, the deflationary consequences, and what policymakers misunderstood about what they were trying to restore

War finance and inflation – the clearest historical cases of the real resource constraint, where the productive capacity question was so obviously primary that the financial question barely arose

Bretton Woods as a hybrid – fixed but adjustable exchange rates, capital controls, gold convertibility for central banks only, and the Triffin dilemma that eventually broke the arrangement

Hyperinflation: Weimar and Zimbabwe – each examined as a specific causal story. Weimar involved reparations requiring payment in gold or acceptable foreign exchange that Germany could not earn through exports, combined with wartime fiscal deficits, exchange-rate collapse, and the Ruhr occupation. It was a confluence of pressures the Reichsbank accommodated. Zimbabwe involved a collapse in agricultural output alongside continued government spending into a shrinking real economy. Neither reduces to “the government printed money and therefore inflation”.

Further themes

The framework also extends to three broader economic debates where precise mechanical analysis is typically absent from the political argument:

Endless growth or de-growth – what GDP growth actually means in Proxy Australia’s accounts, whether it is the right variable to maximise, and what a steady-state economy looks like in the model’s terms.

Migration – simultaneous effects on labour supply, demand, housing, fiscal balances, and money velocity, held together in the model rather than cherry-picked in support of a prior position.

Technology and AI – a productivity shock examined through the model: who benefits, who is displaced, what the fiscal and monetary response needs to do to manage the transition, and why the orthodox view that technology always raises all boats eventually is an empirical claim about adjustment speed rather than a logical necessity.

The article list above represents the current plan; the order and numbering of future articles may be revised as the series progresses.

A short while ago, someone on X asked Grok to explain Japan’s rising bond yields, as if to a ten-year-old. Grok’s answer used the obvious comparison: Japan as a family with the world’s biggest credit card bill, now facing higher interest payments it can barely afford. It’s a natural instinct – households borrow, households pay interest, and households can be crushed by debt. Surely a country is the same, only bigger?

It isn’t, and the reason why needs to be explained properly, because the error continues to do more damage than a single chatbot reply. Every time a national government’s finances get compared to a family’s, the conversation about what that government can and can’t afford starts from the wrong premise. A household cannot issue the currency it owes its debts in, but a monetary sovereign like Japan can. You cannot fail to pay a debt denominated in the thing you alone create, and that isn’t a contestable claim about fiscal policy so much as a plain accounting fact.

Stated outright, though, it can sound like a trick, or an excuse for reckless spending. So rather than asserting it directly, it helps to turn to a comparison almost everyone already has sitting on a shelf somewhere: Monopoly.

The banker isn’t a player

Every game of Monopoly has a banker, and the rules of the game – the actual printed rules, not house variations – include a line that gets skipped over constantly: the bank can never go broke. If the bank runs out of paper money, the players are instructed to write IOUs on plain paper and keep playing. The bank cannot be foreclosed on, cannot miss a payment, cannot be told by a supplier that its cheque bounced. It issues the only money that exists in the game, so there is no one above it to say no.

This is the whole distinction the household analogy erases. A family takes on debt in a currency it doesn’t issue – pounds, dollars, and so on – and can actually run out, because someone else controls the supply of that money and can refuse to extend more. The banker in Monopoly is the equivalent of the currency issuer: the government of Japan, Australia, or the United States, issuing its own currency, borrowing only in that currency, and under no obligation to anyone for the right to create more of it. The players are the equivalent of everyone else – households, businesses, even state governments – who use the currency but don’t issue it, and who can absolutely go bankrupt within the game, in a way the bank never can.

None of that means the banker’s decisions are free of consequence, though – it means the consequences are the wrong ones if you’re still thinking about solvency.

The nature of the real constraint

Open the Monopoly box and count the pieces: thirty-two houses and twelve hotels. That’s it, a hard physical limit built into the manufacturing of the game. Regardless of however much money is circulating around the board and however many buildings players want to build, there is a point past which no more houses or hotels exist to buy. That’s the binding limit in the game, and it has nothing to do with how much cash the bank is willing to pay out.

Early in a game, most properties are unowned and nobody has started building. If the banker pays out generously – big rewards from Chance and Community Chest, everyone collecting on the way past Go – that money mostly goes toward buying up idle property. Nothing was being built before, but now it is. Growth like this is just absorbing spare capacity, and it’s hard to overdo, because there’s so much slack in the board early on.

The trouble starts in two different ways as the game matures, and they should be addressed separately, because they produce different symptoms.

The first is money arriving too early relative to how the game has actually developed. Ordinary purchases from the bank are fixed-price, printed on the board, so a flood of cash can’t bid those up directly. But it does bid up the two channels in the game where price can move: auctions, when a player declines to buy and the property goes under the hammer, and private trades between players. If cash floods in before monopolies have settled through the normal rhythm of buying and landing on squares, those auctions and trades start running hotter, and ownership moves faster and more arbitrarily than the game’s normal pace would produce. Properties end up with whoever has the most spare cash to throw at a trade, not necessarily whoever worked the board best. That’s asset-price inflation: money is chasing a fixed stock of things considered worth owning, bidding up the final price and future valuations rather than creating anything new.

The second is money arriving too late – once most of the thirty-two houses and twelve hotels are already built. In the game itself, rents are fixed on the title deed cards, so Monopoly doesn’t literally model rents rising. However, it does model the constraint underneath that process beautifully: once the houses are gone, extra cash cannot turn into more development because there is nothing left in the box to build. So cash either just accumulates with players, or gets thrown at trades and auctions for whatever scarce capacity remains. In a real economy, where housing, labour, energy and imports are priced continuously rather than printed on a card, that same shortage manifests as rising prices instead of a fixed rent schedule. The limit was the physical count of little green and red buildings in the lid of the box and on the board, not the money supply.

Both failure modes look, from the outside, like “too much money”. Yet, one of them is actually about timing relative to unused capacity, while the other is about hitting a hard physical ceiling. Conflating them is exactly the mistake the household analogy invites, because a household worries about money running out, never about running out of things to buy with it. The houses and hotels represent a real resource constraint – the actual limit on what an economy can do, as distinct from how much money exists to try to do it with.

Players can go broke. The banker cannot.

There’s one more distinction to draw out: the difference between a currency issuer and a currency user. An individual player in Monopoly can absolutely go bankrupt – land on the wrong hotel at the wrong moment, and there may be no way to cover the rent. Players hold a finite stock of the bank’s money, and when that stock runs out and they can’t raise more through mortgaging or trading, they’re out of the game. The bank doesn’t have this problem, because the bank doesn’t hold a stock of money that can run out – it is the source of the money, not a holder of it.

That gap between a player’s position and the bank’s position is similar to the gap that separates a household, or a state government inside a currency union, from a national government that issues its own currency and lets it float. A household holds a limited amount of money at any one time, and depends on someone else – a bank or other lender – agreeing to extend more once that runs out. If they refuse, the household is stuck, and so that is a real balance sheet constraint: a stock that can run out, held on terms someone else can decline. A national government issuing its own currency faces something different, and it has nothing to do with what exists to pay with – it is a limit on what actually exists to buy, the labour, materials, and capacity an economy can put to work at any one time, the real resource constraint instead. Treating those two constraints as though they are the same thing is the entire error built into the household budget analogy.

A policy choice, not a fate handed down

Here’s where the analogy pays off on the specific question Grok was asked. Japan’s benchmark bond yield has climbed to its highest level in roughly three decades, and the reporting behind that move indicates several things at once: the Bank of Japan lifting its policy rate through 2026, concern over the scale of Japan’s fiscal spending plans, a weak government bond auction, and a yen trading near multi-decade lows. That’s obviously a very different story from a household whose credit card company unilaterally decides it is higher risk and raises the rate accordingly. Scott Fullwiler’s research on this is a useful anchor: for a government issuing its own currency under a floating exchange rate, long-term rates are shaped primarily by current and expected central bank policy, with term and liquidity premia layered on top. The size of the debt or deficit matters mainly insofar as it shifts what markets expect that policy to do next, not because lenders can simply refuse to hold the debt.

Translate that into the board game. The banker could house rule the interest rate the bank pays on cash savings – or on any bonds if the game included those – and other players’ appetite to hold Monopoly money (or bonds, if applicable) instead of spending them can still shift what rate the banker judges necessary to keep the game running smoothly. It isn’t a rate the bank negotiates with players who might walk away – rather, if the banker decides to pay more, that’s a policy choice about how much purchasing power to hand to whoever’s holding money instead of spending it. This decision has real distributional consequences, but it is not a sign the bank is under any obligation it might fail to meet.

Raise that rate and more money flows to financial investors – those with cash or bonds. Whether that has real consequences depends entirely on where the game is in its life cycle – early, with houses still to build, or late, with the box already emptied out. Whether the bank can pay isn’t really the issue: a sovereign currency issuer cannot be forced into involuntary default on debt denominated in its own currency, whatever legal, political, or self-imposed rules it may separately choose to operate under. What counts is what that extra money does once it is in the hands of the small subset of players with financial investments – cash or bonds.

Where this leaves the household analogy

None of this means governments face no limits, though it does mean the constraint has been misidentified. The household analogy leads with the wrong issue – where will the money come from – when what is important is where the spending is directed, and whether the game board has the capacity to absorb it. A government spending heavily on imported machinery has a foreign-currency challenge: what it can obtain from outside its own economy, in someone else’s currency, on someone else’s terms. A government spending heavily on domestic labour that’s already fully employed runs into a capacity constraint instead – the Monopoly-houses-running-out problem. Neither of these is an issue about whether the issuer of the currency can be forced into default. It can’t, on debt denominated in its own currency. Instead, what happens to the real economy as a result of the spending is what needs to be addressed.

This piece stands on its own, but it’s also the natural doorway into a longer project I’ve been building: a sequential series constructing a fictional economy, Proxy Australia, from first principles – how money gets created, how government spending and taxation actually work, how banks lend, how deficits and bonds function, and where inflation really comes from. The banker who cannot go broke, and the thirty-two houses that can run out, is the picture worth keeping in mind before any of that is fully built out. It’s also, not incidentally, a picture most readers already have.

This piece isn’t a forecast of where oil prices are heading. Instead, it is an argument about how the market has reached its current conclusion, and why that interpretation is incorrect. Ultimately I may end up wrong about where prices go. However, even if I am, while the market simply chooses to ignore what the fundamentals actually show, I would prefer my error to be for a reason that was demonstrably clear and logical.

The market’s “global glut” argument mistakes a localised bottleneck for a global condition, then treats the price fall that it has caused as confirmation.

Two finite mechanisms are pushing prices down: coordinated strategic reserve drawdowns, and the clearing of the Gulf’s wartime backlog. Both are one-off and both are draining.

UAE and Iraq heavily discounted the trapped oil that exited the Gulf to clear the backlog, and China’s teapots opportunistically bought it cheap. Iran didn’t match that temporary discount, leaving its unsold oil floating at sea while it focused on longer-term buyers.

The pattern of a localised, rather than a global, glut is visible across West African crude too. Those prices have fallen unevenly, rather than in lockstep, reflecting a single mechanism unwinding rather than a global surplus.

A global surplus should broaden across grades and regions, and eventually ease product markets too. What we are seeing instead is concentrated distress in the grades and routes tied most closely to China’s absence and the Gulf backlog.

OPEC output in June remained roughly a quarter below February levels, and US crude inventories are near a multi-decade low. That fits a one-off backlog clearance pushing the price down worldwide while the underlying position stays tight – not a global surplus in the making.

I set out the local-versus-global distinction in detail in Reading the Spread – a price fall can come from a global surplus or from a bottleneck at one point in the chain, and the two are hard to tell apart on a spread screen. This piece updates that argument with what’s happened since.

OPEC output rose 2.34 million barrels a day (mb/d) in June to 18.75 mb/d according to Bloomberg’s survey – still 28% below February once the UAE’s exit from OPEC is factored in. Reuters puts June output higher, at 19.43 mb/d on the same post-UAE basis; measured against Bloomberg’s implied February base, that would make the shortfall closer to 25%. That’s a minor difference in method rather than a disagreement about direction, and both show output well short of pre-war levels.

A local supply problem can still move the global oil price, because the barrels involved are priced into the same global benchmarks everyone watches. However, Brent falling doesn’t indicate whether the fall is local or global. The pertinent question is what is driving it.

For years, Iran’s discount to Brent has really been a sanctions tax, not a market price. Teapot refiners – China’s small independent refineries mostly based around Shandong – often demanded $8-15 off Brent to compensate for the risk of buying sanctioned oil through shadow-fleet, non-dollar routes. The US Treasury’s General License X temporarily removed much of that risk, until 21 August. As a result, Iran’s discount narrowed to just $3 – closer to a normal quality discount rather than a risk discount – at exactly the same time the Gulf’s own discount was widening to clear its backlog. Thus, for a short time, Iran ended up looking more expensive than sellers who had no choice but to discount harder.

TotalEnergies chief executive Patrick Pouyanné commented that Gulf producers are “desperate to sell” crude they built up during the war. However, they face difficulty getting tankers through Hormuz because many shipowners remain unwilling to take the risk, hence producers have been forced to discount heavily to move oil regardless. He expects the market to take three to four months to rebalance.

Iran didn’t cut its price to match the discounted crude, but that’s not really because of stubbornness or inability. It wasn’t short of buyers, and had already moved enormous volumes as positioning rather than spot sales hunting for a home, so chasing a temporary distress discount to win a handful of extra barrels wasn’t worth it. Iran appears to be using the 60-day window to test and rebuild longer-term buyer relationships while dollar settlement is temporarily authorised, rather than winning a short-lived price fight it would later have to reverse.

The result has shown up in the level of Iran’s oil on the water, the shipping industry’s term for crude that’s loaded and at sea but not yet delivered. It has risen to 58-68 mb, according to Vortexa, with more than 20 mb idling in Asian waters for at least a week and over 90% of cargoes showing no confirmed destination.

Most of that ostensibly stranded oil looks deliberate rather than accidental. TankerTrackers recorded roughly 20 mb leaving Iran in a single day on 19 June. Historically, Iran’s exports to China have been around 1.4 mb/d, with reported short-period peaks around 2.2 mb/d. So this single convoy was the equivalent of nine to fourteen days of Iran’s normal sales to China. That looks less like normal spot selling than pre-positioning: either for pre-arranged buyers, for floating storage near Asia, or for optionality ahead of another possible disruption. This fits the earlier point made in Better Flows more reasonably than an account of Iran being caught out by the market. Iran was positioning ahead of a possible new closure of the strait, not caught out by a market it didn’t see coming.

Oil from West Africa tells the same argument from a different angle, and although the pattern isn’t identical for every producer, it illustrates the mechanism described above for the Gulf-China corridor rather than a separate one. During the shortage, Asian refiners paid a premium for West African crude because Gulf oil wasn’t reaching them – any barrel that roughly matched their refineries would do, even at a high price and a less-than-ideal grade. However, with a sudden return of cheap Gulf oil, that premium had no reason to continue for barrels that were only ever a stopgap. Congo and Angola swung most from premium to discount. Congo’s Djeno crude went from a steep discount before the war to a premium during the shortage, and back to a record discount now – S&P had it at $11.86 below Brent on 26 June, and a July cargo offered at $14 below Brent found no buyer at all. Angola followed much the same path: Cabinda and Clov traded $10-15 above Brent in April, and by late June, Dalia, Pazflor and Hungo were selling at their lowest prices on record, with some cargoes offered close to $9 below Brent. Nigeria has moved in the same direction, but by smaller amounts: Bonny Light and Qua Iboe also spiked more than $10 above Brent during the shortage, and have fallen back to modest premiums of $2-3. Nigerian grades are more broadly useful outside Asia and were never bought purely as a stopgap the way some Congolese and Angolan cargoes were, so there’s more underlying demand holding their price up even once the emergency buying stops.

If the world did have too much oil, the size of the price fall in each country would be roughly in-sync, adjusting for things like oil quality and shipping cost, not on how much of that country’s recent demand was really just a temporary substitute for Gulf oil. Yet the latter is exactly what we saw: Congo and Angola, which were bought almost entirely as substitutes, have fallen the most. Nigeria, which has other buyers and uses beyond that emergency demand, has fallen by much less. That pattern only makes sense if this is one temporary cause unwinding at different speeds in each country, not evidence of a real global surplus.

The US SPR has fallen to 326 mb, the lowest since May 1983, and US commercial and strategic crude inventories combined dropped to 743 mb, the lowest since 1984. That is not what a market drowning in oil looks like. If the world were actually oversupplied, the US Government would be buying oil at these prices to rebuild its reserve, not still running it down.

Yet that’s exactly the contradiction in Washington’s behaviour. In 2025 the US had been refilling the SPR, buying at $79 a barrel, or below. That policy was paused on 11 March 2026, when the Energy Department redirected the reserve into releasing 172 mb as the US share of the IEA’s coordinated wartime response – a release scheduled to run roughly 120 days, putting it on track to finish right about now. Prices now are below the $73-79 range the government was previously happy to buy at. If officials believed this was a lasting surplus, the obvious move would be to finish the emergency release and go straight back to buying at an even better price than before. Instead, the reserve is still being drawn down, not refilled. That does not prove officials think the current price is wrong as the release was already scheduled to run its course, but it does sit uneasily alongside the claim that the physical market is already drowning in surplus crude.

Product markets are sending the same mismatch signal from another angle. The 3-2-1 crack spread – the standard measure of refining margin on gasoline and heating oil – has continued climbing through early July. The gasoline calendar spread has strengthened to levels last seen during the acute phase of the war, while the equivalent crude calendar spread has turned negative for the first time this year. Refiners don’t typically earn margins like that in a market drowning in crude, as an actual global surplus would be expected to show up as weakness in both crude and the fuel made from it, not a growing gap between the two.

Morgan Stanley has argued the surplus is here to stay because of what they call two “solvers” – high US exports and low Chinese imports – that let the market absorb the Hormuz disruption without prices spiking as much as feared. The scale involved is substantial: IEA data shows Chinese seaborne imports falling from roughly 14.2 mb/d in March to about 8.6 mb/d by May, while total US oil and product exports rose over the same stretch by about 2 mb/d to nearly 12 mb/d. So it’s accurate in terms of total aggregate oil and product flows, but most of the extra US oil is light and sweet, while the disrupted Gulf oil is mostly medium and sour. These are different grades that most refineries can’t simply swap for each other. Counting US flows as a like-for-like replacement for Gulf crude overstates how interchangeable they actually are. Further, this relies on the bold assumption that lower Chinese imports are a permanent phenomenon, rather than temporary while drawing on reserves – there is no evidence to support this paradigm shift to lower imports.

Goldman Sachs and Citi both expect a surplus by 2027 – Goldman estimates roughly 3 mb/d before global reserves are rebuilt and 2 mb/d after, and Citi has flagged the possibility of Brent returning to $60 by year-end as Hormuz traffic normalises. Worth noting, though, that investment banks have a professional incentive to cluster around similar forecasts, since being wrong together costs a career far less than being wrong alone, so multiple banks agreeing isn’t as strong a signal as it looks. The evidence this piece actually relies on is the OPEC production numbers, global reserve and inventory levels, and the rates of loading, return and transit of tankers, all measured by tracking firms and government agencies, not forecast by banks with a reputation to protect.

What looks like a global glut is actually two finite mechanisms and one open question, not a sustainable surplus. The US strategic reserve releases and the Gulf’s wartime backlog are both one-off and both draining on a timeline that can be estimated, and both are close to their end dates now. China’s absence is doing the rest of the work in making this look like a glut, and its duration isn’t knowable the way the other two are.

OPEC+ approved a further 188 kb/d target increase from August on 5 July – a figure small enough, against total OPEC+ output of around 40 mb/d, to be virtually a rounding adjustment. Yet feeling the need to increase production – even such a tiny one – is not a sign of confidence in a lasting surplus. Even the most recent supply-side headline doesn’t move the physical picture much. On the second part of the race Rystad Energy – an independent energy research and consulting firm – expects about 85% of the lost Gulf production volume back by early in the fourth quarter and all of it by January 2027. That is consistent with the trajectory discussed in Counting Barrels, but those timelines are not guaranteed because recovering production won’t count for much if the oil can’t regularly get out.

Transit through the strait continues to be a useful illustration of how unsettled this situation remains. Total traffic remains well below the pre-war norm of roughly 130-150 ships a day, while commodity-vessel crossings have recovered only unevenly and remain sensitive to which route is being used. That fragility was particularly apparent in the first days of July. Over 3-4 July, at least eight vessels attempting the Omani route reversed course, four of them then shifted toward the Iranian route and exited the strait, before Omani-route traffic partially recovered by the end of the weekend. Nothing about the underlying 60-day deal changed in that window – nobody renegotiated anything, no new agreement was struck, and no official explanation was given. Ships just stopped then started again; the disruption and recovery happened on their own, which is exactly the kind of unpredictable friction market is ignoring on the normalisation side of the race as current prices assume Gulf flows keep recovering smoothly before the backlog and reserves run out.

Iran continues to assert its control over the strait, forcing ships toward its own designated channel and leaving other routes subject to unresolved issues like mine-clearing and security assurances. The question of tolls once the current 60-day free-transit window ends likewise remains unresolved. There’s no agreement yet on what happens once the fee-free period expires, and Washington has previously indicated that paying tolls to Iran’s Government would itself be a sanctionable act, leaving shipowners exposed if Iran revives tolls after the window closes.

If Iran uses that leverage to slow or tax the return to normal flows, the production recovery Rystad projects doesn’t translate into the market normalisation the current price assumes. These uncertainties could result in the backlog and reserves disappearing before real flows are back, opening a gap the market currently has no reason to expect, because it isn’t pricing this as a race with an uncertain second leg at all.

Headline volume figures make that spike look like genuine recovery, but it isn’t that simple.

Most increase was trapped backlog finally exiting rather than a change in the underlying rate.

Mix of vessels crossing has skewed hard toward VLCCs.

Flows quoted are barrels, not ships, so the skew is invisible in the headline number – a handful of oversized cargoes make an exit queue look like a recovery.

Whether there is a meaningful, sustained uplift underneath any of this remains unclear.

Restoration runs through four separate gates, not one, each capable of being the binding constraint on its own: production, return of empty tankers, loading capacity at the terminals, and transit through the strait itself.

China adds a further wild card on top – its own reserves and import behaviour – capable of absorbing or amplifying whatever the Gulf actually does.

Two Spikes Coming made this point only in passing: the market priced the 60-plus vessel day on 24 June as recovery, when it was merely the stranded queue clearing. US Energy Secretary Chris Wright then made the identical mistake just days later, posting on X that “72 ships and 20 million barrels” had transited Hormuz in 24 hours, “fully restoring pre-conflict flows”. Wider commentary drawing on Goldman Sachs’s estimate that Persian Gulf exports had reached 79% of normal levels by late June has repeated the error. Reuters noted at the same time that vessel counts remained a fraction of the pre-war daily average, a discrepancy Wright waved off by pointing to “bigger ships”. In the past fortnight these items have fed a narrative that the strait has substantially reopened and that the supply shock from the Iran war is unwinding faster than anyone expected.

Ironically, Wright’s apparently flippant defence is the whole argument of this piece, offered by the official making the claim it undermines. Goldman’s own estimate of full normalisation, meanwhile, has moved repeatedly in the past three months – end-June, then late August, then end-July, each revision paired with a different price forecast – and at no point has the bank itself declared the recovery complete. The now frequently quoted 79% is just a snapshot from one exhibit in a sequence it keeps revising.

Despite claims of a restoration, total vessel traffic through Hormuz has remained suppressed. Windward’s daily counts through late June and into July put total transits at 32 to 43 a day (with the noted exceptions of 24 and 25 June, at 60-plus), against a prewar baseline that sits somewhere between 93 and 138 depending on the tracker, but clusters most credibly around 100 to 120. That figure covers everything crossing the strait in both directions, from container ships and bulk carriers to LNG and LPG carriers alongside tankers, which account for about half of the total. Even taking the low end of the pre-war range, transit vessel traffic through the strait was running at roughly a third of normal throughout the period described as a near-complete recovery.

Tanker-specific traffic looked far healthier during that peak. LSEG’s data, tracking vessels above 10,000 deadweight tonnes, showed daily crossings reaching around 20 by late June, against a pre-war tanker baseline that clustered mostly in the high-20s to mid-30s.¹ Read on its own, that would put tanker recovery closer to two-thirds of normal. However, the peak in tanker counts coincides exactly with the window the backlog was clearing fastest, which raises the same query about Goldman’s volume figure: is the tanker count itself just another view of the same draining backlog, rather than independent evidence that transit capacity has recovered? Regional export volumes at 79% of normal (Goldman’s figure, which blends Hormuz with the Yanbu and Fujairah bypasses) sitting alongside total Hormuz vessel traffic at roughly a third of normal are not measuring the same system, but the gap between them is still telling: it looks much more like a temporary change in mix than a change in rate – a reflection of the unusual circumstances created by so many vessels being trapped in the Gulf at once.

The MOU did not release a normal, or even a random, sample of suppressed traffic. It released a queue that had spent months loading, skewed toward VLCCs, each carrying close to two million barrels, displacing smaller vessels that carry a fraction of that. VLCCs cluster in a release for a mechanical reason as much as a commercial one. A supertanker takes the better part of a day to load even at full pumping rate, so at any moment during a closure a disproportionate share of the vessels caught mid-cycle are VLCCs, simply because they spend longer in that stage than a smaller tanker does. So once the strait reopened, there would be at least double the normal daily VLCC count.

Consider a deliberately simplified model of that spike period. Granted, this could be calculated in more detail, but the point is illustrative and directional, so the aim is to be roughly right not precisely wrong.

Run that same 62% of ships through the ordinary, unskewed pre-war mix instead, and volume would also land at 62% of baseline, since volume scales directly with ship count once the mix stops changing. The gap between that mechanical 62% and the modelled 93% – something over 30 percentage points – is what the VLCC skew alone is contributing, on top of whatever the ship count itself is adding.

The 22-ship figure for 24 June lines up with what Windward recorded: outbound-only counts around 41 ships that day and 45 the next, once inbound ballast crossings are subtracted from the reported daily totals, with half of that range landing almost exactly on the model’s 20 to 22. Tankers made up roughly half of all outbound traffic during the peak, consistent with a release skewed hard toward the vessels that had spent months loaded and waiting. By the end of June, total outbound traffic had roughly halved from that peak, though whether tanker-specific traffic followed the same path remains unclear either way.

Rory Johnston’s (RJ) recent tracking describes Hormuz recovery – inbound ballast especially – as running ahead of his own expectations, yet his own chart dips on several recent days, worth treating cautiously rather than as a trend, given how sensitive single-day readings have already proven to be across every tracker observed. LVision_Trading’s (LVT) daily transit-capacity estimates – on a crude-only basis – confirm the same instability during the same window. Flows vary between above normal and roughly half of normal within the space of a week or two, driven alternately by queue-clearing surges and by Iran’s demonstrated willingness to reintroduce risk the moment it judges the terms of the MOU breached. A rate this volatile has no single value that can really be quoted as “the recovery level”, and every headline percentage drawn from it – investment banks included – is a snapshot of wherever flows happened to be on the day they were measured, not where it has settled.

The backlog evidence actually confirms this, independent of any hull count. According to Kpler, the pool of loaded tankers ready to depart was 118 in mid-June. HFIR’s broader estimate of the total crude trapped in the Gulf was roughly 170 million barrels. RJ separately tracked the oil sitting on ships in the Persian Gulf falling from about 160 million barrels at the start of June to around 110 million by 25 June, a decline of 50 million within three to four weeks. The sum of daily outbound volumes tracked over six days around the Ever Lovely and Kiku incidents is roughly 105 million barrels across 22-27 June, and the portion of that falling before 25 June, roughly 51 million barrels, closely matches RJ’s figure. Granted, the observations are not measuring exactly the same thing – RJ’s is a net stock change over three to four weeks, changing with loading, discharge and reclassification as well as departures, while the outbound estimate covers a narrower window of cargo actually leaving – so this is not meant to be a precise reconciliation. Yet there is a consistency in order of magnitude which does not contradict the argument presented.

None of this is a criticism of either tracker – two careful analysts can measure differently defined things and land on different numbers without either being wrong, especially with an unclear dataset.

A majority of current Hormuz traffic, by some estimates as much as 57% to 65% is transiting dark to begin with, which means every tracker is making its own modelling choice about how much of that unobserved activity to add back in. AXSMarine’s own tracker states outright that its figures “may understate actual volumes” for exactly this reason.

Part of what any of these trackers are picking up may also be a specific commercial operation rather than a broad signal of Gulf-wide capacity returning. Sinokor Maritime, a Korean shipowner, began an unusually large VLCC-buying campaign in December 2025, months before the war, and by January had bought 35 of the 45 VLCCs sold globally that month. Once Hormuz reopened, the fleet was reportedly put to work on a repeated shuttle: load at a UAE terminal, transit Hormuz, transfer cargo at Fujairah, return, on a cycle of roughly five days, replicated by Turkish and Greek owners at freight rates several times the pre-war norm. LVT estimated the pool at 25 to 40 vessels, mostly VLCCs, capable on its own of 8 to 12mbd of shuttle capacity. A dedicated fleet running fast, premium-rate round trips would show up in transit counts and in VLCC-heavy volume readings without saying much about whether the wider Gulf’s transit capacity has actually returned.

The UAE supplies a clear illustration of the argument. Crude production reached 2.8mbd in May, the IEA reported, still about 835,000 barrels below pre-war levels – roughly 77% recovered. A separate reading, from Trading Economics using OPEC secondary-source data, suggests May production was closer to 2.1mbd, roughly 62% of the same baseline. The two disagree by more than 700,000 barrels on the same country in the same month. The uncertainty so far identified has been tracker disagreement about how much oil is moving: vessel counts, ballast crossings, and transit volume. This one is about how much oil is actually coming out of the ground in the first place, and even that isn’t settled between two credible sources, and being misrepresented in media commentary.

Even taking the higher of the two figures, production alone does not support media headlines. Crude and condensate exports rose from 3.1mbd in May to a record 3.7mbd in June, according to Kpler and Vortexa – comfortably above the UAE’s own 3.1 to 3.3mbd pre-war range, and achieved, on Kpler’s own account, through a mix of resumed Hormuz transits, dark-vessel shuttle runs with transponders switched off, and “unwinding part of its inventories”. None of that is production, though – it is the same argument expressed above. Stock built up during the closure has been sold down once the strait allowed it and OPEC no longer caps it, the UAE having left the group on 1 May. Reporting record exports as a restoration of flows and production masquerades as recovery, when it was mostly a queue clearing rather than a return to normal. That is exactly what the transit-wide numbers in this piece have been arguing, now visible inside a single set of national accounts.

Whether the next few weeks bring a plateau, a continued climb, or another reversal is impossible to say with confidence right now. That does not mean the strait is as closed as it was in March-May, but a fortnight of elevated volume readings, and the tanker hull counts moving alongside them, produces limited evidence about the sustainability of the strait’s throughput, whatever the headline percentage suggests. A count of ships proves no more than a count of barrels, when both are drawn from the same short window a queue happened to be clearing through. And the ceasefire underwriting all of it remains unsettled: talks in Doha are paused until after this Saturday’s funeral for Iran’s former supreme leader, Iran’s army issued a fresh warning against any US “interference” in the strait, and Oman’s proposal for transit service fees is still opposed by Washington and unresolved.

There is a larger reason for caution underneath all of this, one that will eventually need deeper treatment than a paragraph here. Chinese seaborne crude imports have fallen well outside their own five-year range through 2026, down from a 2025 monthly average above 10.6mbd to 7-8mbd in recent months, even as Middle East crude sitting onshore and afloat has climbed by more than 47 million barrels since January. RJ and JPMorgan have separately described the result as a mini-glut: barrels arriving faster than the market’s largest buyer wants them, which is a more plausible explanation for recent price weakness than any genuine restoration of flow.

Restoration ultimately depends on four separate gates, not one: production, the return of empty tankers to collect it, loading capacity at the terminals, and transit through the strait or its bypasses. Each is a separate point of failure, and restoration can only move as fast as the slowest of the four.

Wright’s claim covers the last gate alone, and an inflated reading of even that. Goldman’s 79% is broader, blending several gates together, but it still does not explain how much of the observed flow is current production and loading, against how much is simply stock that had already been loaded and was waiting to clear.

Production should be the clearest of the four gates to measure, which is a low bar, but even that is unclear. Kuwait’s state oil company put output at roughly 73% of pre-war levels by early July, already ahead of its own forecast of 70% within six to eight weeks of reopening. ADNOC had forecast 80% for the UAE by October-November. The UAE’s own current production figures are contested, as set out above, but even the more generous reading leaves it short of pre-war levels, sitting somewhere between 62% and 77%. Taking Kuwait and the UAE together as a rough proxy for the wider Gulf, production alone still implies a meaningful shortfall against pre-war levels, even once transit stops being the binding constraint. Saudi Arabia remains the open variable that could push the true figure either way.

Loading is bounded by terminal capacity that was never built for a full surge across every Gulf producer at once. And reserves and inventories – the Strategic Petroleum Reserve and Cushing most acutely, where further large drawdowns become progressively harder to sustain – are outside those four gates entirely. They are the buffer that has let apparent supply to consumers exceed what production, ballast return, loading and transit could sustain on their own, and that buffer is what runs out first if the four gates don’t catch up.

The same race laid out in Two Spikes Coming – between reserves and restoration – continues. Nothing in the transit volume data reviewed here changes the odds in that race. If anything, a market that has spent the period since the MOU mistaking a queue for a recovery is a market poorly positioned to notice when production and transit finally settle into their post-war rate, once the backlog has finished clearing.

¹ Pre-war tanker-specific baselines vary meaningfully by source and by the deadweight threshold used. The IEA’s own country-by-country breakdown for full-year 2025, based on Kpler data, puts total Hormuz volume at 19.87 million barrels a day – 14.95 crude and condensate, 4.93 products – close to the 20 million barrel round figure used in the model above, and the basis for it. Vessel counts specifically are less settled, and depend heavily on whether a chart is showing one direction or both, and which vessel types it includes. LSEG’s own bidirectional series suggests an outbound-only figure closer to 15-20 tankers on a typical pre-war day, but these are only larger tankers. A separate PortWatch series, which appears to track total vessel arrivals across all cargo types rather than tankers specifically, implies a tanker-only figure somewhere in the 35-55 range once other vessel types are excluded. Neither is a clean, direct read. The illustrative model above uses 30 as a rough midpoint between these disagreeing estimates, not as a figure drawn precisely from any single source – treat it, like every other vessel-count figure in this piece, as indicative rather than settled.

Iran is entrenching control over the strait through mechanisms – mines, fees – that outlast any ceasefire.

Iran’s institutions can’t agree among themselves, so even a signed deal still may not compel the IRGC, hence the physical reopening the market is pricing isn’t coming on the MOU’s own timetable.

Meanwhile the price is being held down by three cushions – released barrels that had been trapped in the Gulf, SPR drawdowns, and Chinese reserves and reduced imports – that are all finite, so the mispricing identified in Two Spikes Coming hasn’t resolved.

New evidence this week – inbound tanker numbers, floating storage, operator testimony – confirms the physical picture rather than the price picture.

The argument in Two Spikes Coming rests on a race, between stockpile depletion and production restoration, with Cushing already near its operational floor and reserves elsewhere running out within one to two months. The past week has not changed that race, but it continues to indicate the reserve draws are still outpacing the return of flow.

The MOU signed on 17 June was supposed to settle the reopening. Instead it has settled into a pattern of brief traffic windows followed by a strike, a US response, and a return to the negotiating table to manage the aftermath. The Ever Lovely was hit on 25 June inside the safe corridor the International Maritime Organisation (IMO) and Oman had set up along the Omani coast days earlier. The US struck Iranian missile, drone, and radar sites the next day. A projectile then hit the tanker Kiku and Iran fired at US positions in Bahrain and Kuwait. The US next expanded its target list to surveillance, communications, and minelaying infrastructure. Both sides then turned up in Doha this week and kept communicating, which suggests the ceasefire itself is intact, but does not indicate the deal underpinning it is progressing.

The two sides are not talking directly to each other in Doha. American and Iranian delegations are meeting Qatari and Pakistani mediators separately, a step back from the direct sessions held in Switzerland two weeks earlier. Iran’s stated priority is Clause 11, the release of frozen assets, and President Pezeshkian has stated that $6 billion of the $12 billion held in Qatar will be returned, though it is not yet clear on what terms or even whether the funds have moved. Iran will not discuss its nuclear programme until those funds move, but if that money moves without a matching concession on enrichment it will reduce the leverage the US has left for that discussion. Trump has claimed a deal was close at least 38 times between late March and early June, according to a CNN count. Doha is yet one more round in that pattern.

The strait’s governance is where the deal has stalled most. Iran’s foreign minister Araghchi has said the removal of “obstacles” in the strait, and its reopening, rests with Iran alone. The IMO’s Secretary-General has said Iran laid an estimated eighty mines across the main shipping channel. The timetable for clearing them is set by Tehran regardless of what Doha produces. The Joint Maritime Information Centre raised the strait’s security threat level to “substantial” this week, citing mine risk and clearance uncertainty. Oman has separately delivered a service-fee proposal to Washington and its allies. An Iranian official has called the fees mandatory, but a regional diplomat has called them voluntary. Either way, Iran can prioritise the shippers who comply and delay the ones who do not, so the dispute over wording matters less than the authority it establishes.

Inside Iran, more than sixty of the Assembly of Experts’ roughly 88 members signed a statement on 28 June warning negotiators against crossing Khamenei’s red lines, control of the strait among them. The Assembly’s own secretariat publicly distanced itself from the statement within hours, which means even the body meant to speak for Iran’s clerical establishment cannot agree on how hard a line to take. Pezeshkian spent the same week in Qom telling senior clerics the opposite, that the MOU was an economic win worth defending. And while the president was making that case, the IRGC struck a vessel inside a corridor the foreign ministry had just endorsed. Three arms of the same state, pulling three different directions, in the same seven days. No single part of the Iranian state can bind the others to one position, hence why incidents the negotiators did not authorise keep recurring.

WTI is trading around $70, close to its level before the war began, and Morgan Stanley has cut its Brent forecast on the basis that Hormuz is reopening faster than expected, projecting a 2027 surplus of 4.8 million barrels a day. “Strip away the narrative,” the bank’s analysts wrote, “and read only the prices. They describe a market that has weakened across the board.” Morgan Stanley may be right about a near-term glut – outbound cargo has genuinely surged since the MOU – but the mistake is extrapolating that burst into durable recovery. The analysts are getting the direction backwards because a weak price does not necessarily prove a weak market. Rather, in this case it means a market distorted by reserve draws and supply disruption, and both of those aspects are temporary props under the market, not foundational features. That makes Morgan Stanley’s case much harder to sustain.

Roughly 170 million barrels of crude that had been trapped in the Gulf cleared the market once the MOU allowed it out. The SPR is drawing at a pace that leaves perhaps three to six weeks of room. Chinese crude imports have fallen by something like 5 mb/d since March, while China’s visible commercial stocks have barely moved, which means the shortfall is being met from reserves that do not appear in any published series. Cushing itself fell to 18.96 million barrels in the week to 19 June, the lowest since October 2014 and near the roughly 20 million barrels traders treat as an operational floor. A partial reopening of the strait does not fix that on its own.

There is limited evidence of a return toward normalised transit through Hormuz that would fix this. Tankers leaving the Gulf are just running down a backlog that had built up over four months, while empty tankers still aren’t returning to the Gulf in sufficient quantities to restore normal export flows.

NYK Line’s chief executive said this week that the strait remains nowhere near prewar conditions. Iranian crude is meanwhile building up in floating storage off Singapore, which looks less like a bet on higher prices and more like preparation for the strait’s next closure.

The race described in Two Spikes Coming has not been decided – it is still awash with uncertainty and chaos. But the reserves are still drawing down, and the flow that would slow that draw still has not been restored.

Cushing, Oklahoma is the pricing point for West Texas Intermediate (WTI) crude and the physical hub through which US oil supply flows to refineries across the Midwest and Gulf Coast. As of 25 June, inventories have fallen to 19 million barrels, below the operational minimum (~20mb) that the industry considers the threshold for physical stress. The US Strategic Petroleum Reserve (SPR) has fallen to 331.2 million barrels, the lowest since 1983. According to the IEA, global inventories are at their lowest seasonal point in recorded history.

The market has priced almost none of this.

What happens with Cushing below operational minimum?

When inventory falls below the operational limits (~20mb) – which Cushing has now breached, at 19mb – the hub enters a stress zone rather than hitting a cliff edge. The operational minimum is not a precise failure point; it is the threshold below which there is no buffer left to absorb further stress. The immediate effects are mechanical.

Pipelines that draw from Cushing operate on pressure differentials, but inventories at this level do not cause pipelines to fail instantaneously. What deteriorates first is scheduling reliability, the ability to guarantee volume and timing. Refineries that depend on Cushing lose the flexibility to absorb disruptions or respond to demand spikes. They are forced to draw on their own tank inventory as a buffer, but those buffers are also depleted.

Blending flexibility deteriorates with it, because the crude that remains at tank bottoms is the lowest quality and pulling it through creates blending problems. Cushing’s core function is blending different grades from the Permian, Bakken, and Canadian streams to meet pipeline quality specifications and refinery feedstock requirements. When you’re scraping the bottom, blending optionality narrows significantly, so refineries either accept degraded feedstock or reduce runs.

The financial effect then follows. WTI physically settles at Cushing, and when physical delivery becomes constrained, physical tightness increasingly dominates futures pricing. You get extreme backwardation and basis blowouts as buyers scramble for physical barrels that can’t be reliably delivered. The WTI-Brent spread would likely widen – the US price and the global price decouple – and with US export capacity under pressure, nothing is left to hold Brent down.

Midwest refineries are the most exposed because they are dependent on pipelines from Cushing and have limited alternative supply access. Gulf Coast refineries theoretically have more flexibility as they can draw waterborne imports, but they are already running at record utilisation. US crude exports have run at record levels since March, with the Gulf Coast processing additional crude to produce the diesel, jet fuel, and gasoline that Middle Eastern and Asian refineries could no longer supply at normal volumes as Gulf feedstock flows through Hormuz collapsed. That additional throughput has been draining Cushing. The crude draw and the product export are the same event viewed from different ends of the supply chain.

The broader consequence is that Cushing operational stress would convert a global supply disruption into a US domestic supply disruption simultaneously. The US has been acting as supplier of last resort to the world because its own inventory position appeared comfortable. Once Cushing signals stress, that export capacity comes under pressure because domestic refinery supply tightens. The price signal that has been suppressed by reserve draws then has to clear through the crude price instead, spiking rapidly.

That is the first spike.

The second follows when the global restocking wave arrives. Most major economies have been drawing strategic reserves throughout this crisis – the IEA coordinated a 400 million barrel release across member nations, Japan and South Korea have drawn their modest reserves steadily, and Europe has done the same to varying degrees. China’s position is opaque – Beijing does not publish reserve levels transparently – but the working assumption is that Chinese reserves have also been drawing to keep refineries running through the Hormuz disruption, given that China normally takes roughly 20% of Gulf crude and its total imports are well down.

When those draws exhaust, every country needs to restock simultaneously. They will all be competing for the same barrels at the same moment, in a market where Hormuz throughput remains well below pre-war levels and the reconstitution sequence – demining, tanker repositioning, hull fouling, refinery feedstock switching – has barely begun. Reopening is not restoration, but the market continues to treat it as though it were.

The shape of the financial fallout is therefore probably two distinct repricing events rather than one continuous move. The market has been reactive and consistently wrong in the dovish direction throughout this crisis. It priced the Islamabad MOU as restoration. It priced the 70-vessel day on June 24 as recovery, when it was just clearing the stranded queue backlog. It will price the first spike as the peak. Then the restocking demand will appear and the market will price that shortage separately, when it can no longer be ignored.

Two spikes. Two moments of forced repricing, each preceded by the same premature relief. Brent at $70 tells you the market thinks the crisis is over. But it isn’t. Brent at $70 instead tells you the market hasn’t looked at what comes next.

There is a law – Murphy’s is the one usually cited, though the universe operates several variants – that guarantees the market will move against you the moment you publish. I spent the better part of this week calibrating Ending the War on Paper, arguing that Brent’s forward curve remained in backwardation, the physical market’s signal that the deficit premium had not gone away whatever the paper market was doing, and that the wave was still coming. Within an hour of publication, the curve flipped to contango.

For readers who don’t follow the mechanics of futures markets, backwardation means prompt crude costs more than future delivery – the signature of a market short of supply right now, as it has been virtually since the war began. Contango is the reverse, with future delivery costing more than prompt, and it conventionally signals that near-term supply is ample, storage is filling, and the price decline reflects fundamentals rather than sentiment. If contango is the right interpretation, the market has correctly priced a surplus, the wave has been cancelled by the flood, and my argument is wrong.

So is it?

Markets are not the font of all knowledge, and price is evidence rather than verdict. We have seen this in enough bubbles and dislocations – episodes where price and underlying reality diverged for long enough to destroy people on the right side of the argument – to appreciate the distinction. Keynes observed that the market can remain irrational longer than you can remain solvent, a point that Christian Bale’s character in The Big Short came uncomfortably close to proving. It is a reason for humility about timing rather than a reason to abandon the analysis. The question is which fundamentals the price is reflecting, and whether those are the ones that determine where oil goes from here.

Two conditions, one signal

Contango can arise from two very different conditions. The first is oversupply across the whole system: inventories building everywhere, buyers with ample stock, sellers unable to clear at spot without discounting into full tanks on every continent. The second is a localised bottleneck – barrels accumulating faster at a specific point in the delivery chain than the receiving infrastructure can absorb, forcing sellers to discount to find the few buyers with space left, even as the wider balance remains tight. On a spread screen the two look the same, but they are not as one is persistent and the other is temporary.

What has happened at Hormuz over the past week is the second condition – the temporary one – and the vessel data shows why. Flows through the Strait since the MOU have reached 10-15 mb/d, and the market has read that as normalisation, but it is not. Those flows are overwhelmingly pre-loaded vessels clearing a backlog, reflecting only the safe passage they needed, not restored production or refilled loading cycles.

Gulf loadings are the crucial variable to monitor. Ballast entries into the Gulf, the leading indicator for new cargo capacity, fell from 20-30 mb/d equivalent in February to near zero through March and May, and have only begun recovering toward 5-6 mb/d on a 10-day rolling average. Roughly 160 million barrels were sitting on ships in the Persian Gulf at the start of June and around 110 million remained as of 25 June. When that floating storage empties – weeks away at current clearing rates – fresh loadings at sub-6 mb/d cannot sustain anything approaching current throughput. The market is watching a queue drain and erroneously calling it a recovery.

Commercial transits on a 24-hour rolling basis stood at 44 as of the evening of 24 June, against a pre-war baseline consistently above 120. A VLCC booked this week for the Persian Gulf–India route cleared at 897 Worldscale, nearly nine times the benchmark rate, the highest of the year. Those numbers are difficult to reconcile with a clean global-surplus story, but they are consistent with vessel shortage and congestion at the exit end of a waterway shut for four months. Hull war insurance premiums have fallen from around 5% of vessel value to around 2% over the past six days, a reduction in perceived risk – though at still roughly five to ten times pre-war levels, 2% signals a waterway in transition rather than one that has normalised.

Asian refiners have spent recent months sourcing replacement crude from the US, West Africa, and Russia, booking forward under term arrangements to cover the shortfall. By the time the MOU was signed, they were covered through August, so their forward purchase programmes are now full, even though global inventories are not. The flood of Gulf barrels now exiting Hormuz is therefore coinciding with a market that currently cannot absorb prompt deliveries quickly, however willing it might be to take September cargoes. Sellers need to move these barrels quickly, and with producers across the Gulf competing for the same buyers, the only lever is price, so they discount. But the discount reflects a shortage of available tank space and unfilled delivery slots, not a shortage of demand for oil. A buyer willing to take a September cargo would pay close to full price, but a buyer being asked to take a barrel now, into tanks recently filled, needs to be compensated. That inconvenience discount is a logistics price, not a market-clearing one.

The crack spread data confirms the same argument from the refinery end. With contango, discounts, and Brent below $74, crude appears weak, yet refined products are not. US gasoline crack spreads are testing their 2022 all-time seasonal highs and diesel is still well above its five-year range despite easing from its war peaks. The simple crude-surplus story does not explain those margins by itself – cheap feedstock should flow through into cheaper products, but it hasn’t. So it is clear that refineries are not the constraint and demand for fuel is not the problem. Instead, the spread is reflecting either or both of two issues. The first is the local storage problem discussed above – crude is backing up in front of processing capacity that cannot absorb the pace of arrival, because suddenly too much has arrived too quickly into schedules that buyers have already filled through different sources. The second is a mismatch between crude now exiting the gulf and refineries not best configured to run that type. They are logistics and grade-routing problems, not a demand problem, reflecting the two-track divergence that Ending the War on Paper argued the market had not priced, but now apparent in the margin data.

The grade displacement can be seen in discounts against crude from alternative sources. Angolan crude is trading at the largest discount in more than a decade, at times approaching ten dollars below Dated Brent, while the North Sea Brent benchmark has swung from a record premium in April to a discount against futures. Angola’s recent buyer, principally China, is shifting back toward Gulf sour crude as it returns to market and the discount reflects losing that buyer to a cheaper alternative rather than a world short of demand. Grades that captured Asian demand during the closure are being undercut by returning Gulf supply and finding new clearing destinations in Europe. The financial press has misinterpreted these moves as a verdict on the global balance.

What the buffers say

Oil is back near pre-war levels, but meanwhile:

• Production remains well below pre-war.

• Transit out of the Gulf is well below pre-war.

• Transit in – empty vessels returning to load – is even further below pre-war.

• Reserves and inventories are well below pre-war.

Markets have been wrong before. Many times.

Crude oil inventories are below the five-year range – not merely below average, but outside the historical envelope, while Cushing is near its operational floor. The Strategic Petroleum Reserve (SPR) has fallen from roughly 580 million barrels before the crisis drawdowns began to around 350 million barrels, and total US crude stocks including the SPR fell to their lowest level since 1984. Commercial crude and product stocks – stripping out propane, propylene, and the SPR to get the most operationally relevant measure – are tracking toward their lowest level since at least 2016 by end of July, the projection sitting below the entire historical range shown on the chart, including the COVID trough. The historical relationship between this measure and Brent is unambiguous: when inventories fell to comparable levels in 2022, Brent was above $100. The current price implies the market has decided that relationship no longer holds, with no evidence to support its determination.

US refinery demand dropped sharply during the war but has recovered toward the top of its five-year range, meaning domestic consumption of crude is robust even as export demand softens. US oil exports spiked to record highs as America became the supplier of last resort, which is why domestic inventories drew down so hard and why the export arbitrage is now closing as Gulf crude returns. These readings show a system that absorbed the Hormuz shock by consuming its buffers rather than finding alternative equilibrium, and those buffers are now thin.

On 11 March, IEA member countries agreed to make 400 million barrels of emergency reserves available to the market – the largest collective action in the agency’s history. That release is finite, and as it winds down toward the end of the third quarter, the stocks drawn from government and industry obligation reserves across Japan, South Korea, and Europe will eventually need to be rebuilt, in part through open-market purchases. OECD inventories are on track, on the IEA’s own June forecast, for their lowest level since 2003 by December.

A market generating those readings is not in surplus. It has a temporary prompt glut at the receiving end of a very long supply chain, sitting on top of depleted buffers at the consuming end.

The dark transit question

Some volumes had been moving through dark transits and the UAE’s bypass pipeline before the MOU was signed, so the net new supply entering official visibility is smaller than the gross backlog suggests. How much smaller is uncertain, and the strongest claims about pre-MOU normalisation trace back, directly or indirectly, to US official statements and CENTCOM data – sources with clear institutional incentives to present the reopening as close to normal. The same epistemological caution flagged in Ending the War on Paper applies here: independent vessel-tracking data and official pronouncements are not telling the same story, and the tracking data is the more reliable anchor. Although there is net new supply, its scale should be held with considerably more uncertainty than the financial press has applied to it.

China

The China picture is more complicated, and the bearish case has harder evidence here than elsewhere. Anecdotal trader reports indicate teapot refineries running at utilisation rates near nine-year lows, and of some Chinese refiners offering cargoes for resale rather than absorbing them, suggest demand weakness that goes beyond the receiving-end bottleneck. If that reflects something durable – refineries operating at structurally lower rates or consumption patterns permanently shifted – then the weakness extends well beyond the Hormuz normalisation timeline. If it reflects temporary feedstock disruption, suppressed margins during the war, and deliberate inventory drawdown as a policy response to high prices, then Chinese demand returns as Gulf crude falls and feedstock costs normalise. The data is not yet conclusive either way, and the wave argument depends partly on a variable that cannot be resolved from current information.

The option Iran is preserving

TankerTrackers data shows Iran exported approximately 40 million barrels of crude between 15 and 25 June, with 20 million of those moving in a single day on 19 June – several days before the formal US sanctions licence took effect. That is not the pace of a party easing cautiously back into the market.

The IRGC has since declared the new Omani shipping corridor “unacceptable and perilous”, and analysts tracking the Strait note that the longer transits continue on the Oman lane without Iranian enforcement, the greater the pressure on Iran to escalate in order to reassert control.

Iran need not be acting in bad faith to behave this way. It is simply acting rationally under uncertainty, as any party would when negotiating with a counterparty whose principals were saying opposite things in public throughout the talks, under an agreement whose central precondition – an end to the war in Lebanon – remains unmet. Iran is shipping barrels now in record volumes, to bank oil export receipts regardless of what happens next, while the precondition remains on the record, available as a basis for reasserting closure if the wider negotiation fails.

At $73 Brent, the price does not appear to be carrying any of that residual risk. The supply now pressing on the contango may prove shorter-lived than the market is assuming, because the party that opened the waterway has arranged the agreement to preserve the option of closing it again, and the pace at which it is moving oil suggests it is not counting on the full 60 days.

What the signal means

The contango tells you about the cost of moving and storing oil right now – where prompt barrels are in excess of what the immediate receiving infrastructure can absorb. But it is not a window into the fourth quarter, when the emergency release winds down, when reserve rebuilds begin competing for the same barrels, when resumed Hormuz shipping must deliver current production rather than clear a pre-loaded backlog, and when the Lebanon question either resolves or doesn’t. The argument in Ending the War on Paper was about that sequence, and the contango flip, however well-timed to cause maximum inconvenience, does not address it.

What the curve is measuring is a queue, not a surplus – and the queue, unlike a surplus, has an end.

The market sold the war premium out of oil within a week of the Islamabad Memorandum of Understanding (MOU), and it was logical to do so. Brent fell from the low $90s to the mid $70s, giving back nearly all the gains accumulated since the conflict began on 28 February, on the reasonable judgement that a signed deal and a lifted US naval blockade would open the Strait of Hormuz and get barrels moving again. The premium for that risk should fall on a credible de-escalation signal, and it has.

This is a complex issue, however, with multiple moving parts, and the market’s error lies in what it has also priced. Oil has faced two premiums since March, but the market has treated them as one and removed too much. The war premium has gone, correctly, but the deficit premium has not – nothing in the MOU increases the volume of oil moving through Hormuz today. Global consumption has run ahead of production for months, the gap covered by drawing down strategic and commercial inventories rather than by replacing lost supply, and a signed deal does not address that – only a return of physical flow does. The forward curve confirms the distinction that the spot price recognises and the market has blurred: Brent is still in backwardation, the signature of a physically tight market, rather than flattening as it would if supply had actually loosened. The paper market has moved lower yet the physical market has not followed it.

That gap will close eventually, and the size of that mispricing gap is what this piece calls the wave: not a new shock, but the delayed recognition of conditions already in place. As argued since April, what the Strait’s closure produced was a hydrocarbon shock rather than just an oil price shock, and the market still does not acknowledge that the two behave differently, which has downstream implications. An oil price shock moves along the supply and demand curve, whereas a hydrocarbon shock operates on two tracks simultaneously. The price track is familiar – energy costs rise, producing a dilemma for central banks. The physical track is the one the market has not priced: the closure restricted the supply of industrial feedstocks – naphtha, LNG, helium, chemical intermediates – that underpin global supply chains with limited substitutes. That restriction does not respond to price signals because it is a volume problem rather than a cost problem. Higher prices do not produce more molecules. The wave, when it arrives, will reflect both tracks – not just the crude price, but the downstream cost of months of petrochemical feedstock disruption working its way through input-output chains that have already exhausted their buffers.

Reopening is not restoration

The political picture is as contested and unclear as the physical one. Washington claimed the Strait was open, but Iran announced a fresh closure on 20 June in response to continued Israeli strikes in Lebanon; CENTCOM denied that control had changed and stated that 55 merchant ships had transited that day. All three are official pronouncements, but should be read cautiously.

Independent vessel-tracking data is a more reliable anchor than political statements. Transits are well below pre-war levels given the Strait is only just beginning to open. Implied flow has slowly ticked up to a 7-day moving average of 8.5 mb/d by 23 June against a pre-war level of around 20-21 mb/d, roughly 40% of normal. Prior to the war, the Strait averaged around 125 vessel crossings per day; this collapsed to single digits from late February. Kpler reported 71 transits over 19-21 June with a weekend peak of 35 on 20 June, and 39 on Monday 23 June. The initial surge reflects pre-positioned vessels clearing – laden tankers that needed only political clearance to move – rather than a restored loading and shipping cycle. Empty tankers returning for new cargoes are not yet visible in the data.

Reconstituting the tanker fleet could take eight to ten weeks before meaningful new supply reaches its destination. Part of the fleet that serviced the Middle East-Asia route has spent more than three months redeployed onto longer routes. Vessels that spent months at anchor have accumulated hull fouling that reduces speed, and the attention they need will compete for drydock capacity with ships requiring war-damage repairs and insurance surveys. Demining in the Gulf remains a work in progress, and shipowners continue to exercise caution in the uncertain environment amid still-elevated insurance premiums.

Mark Lashier, Chief Executive of Phillips 66, explained the next constraint: between 90 and 100 million barrels remain trapped in the Strait, clearing only as fast as onshore storage can receive them. “Most of the tanks on shore are full before crude can appreciably ramp up,” he said. In other words the bottleneck is storage, not the waterway. And even when tanks drain, Asian refineries that spent months running mismatched alternative feedstocks – US light tight oil, West African crude, Russian Urals – do not switch back to Gulf sour crude instantaneously. They need to variously clear feedstock inventories, renegotiate term contracts, physically adjust crude slates, ramp-up production, and manage deferred maintenance. Thus the sequence from an open waterway to lower fuel prices faces at least three transitions – waterway to storage, storage to refinery, refinery to product – each with its own lag. Lashier’s conclusion: “There’s going to be some structural shift in what the crude floor is”.

The shortfall from disrupted Gulf exports was roughly 11-13 million barrels a day at its peak. Before the war the Strait carried around 20% of global petroleum liquids, roughly a quarter of seaborne oil trade and around a fifth of global LNG trade on top of that. The IEA puts full production restoration at January 2027, while ADNOC’s Group Chief Executive Sultan Ahmed Al Jaber estimated flows at 80% of pre-war levels no earlier than four months from an effective agreement. That suggests around November 2026, with full normalisation not expected until Q1-Q2 2027. US shale or other alternative sources cannot fill the gap. Output has barely moved from its pre-war level, and energy companies are unlikely to make the necessary long-term investment commitments in such uncertainty. Further, equipment and workforce reductions from the pre-war low-price period cannot be reversed quickly, and new production takes six to twelve months to crystallise.

Buffers and China

Strategic reserve and commercial inventory drawdowns have absorbed the supply shock, but only temporarily. Even on a clean and rapid reopening of the Strait, reserve draws will slow before both production and shipping flows have returned to pre-war levels. The lag between those two conditions could run to months, reflecting bottlenecks in storage and at refineries. There are several difficulties in forecasting the length and size of this lag. There is the production side, discussed above, and also the impact of strategic reserves given lack of past experience to draw on. This reflects determination of the appropriate floor, at which point drawdowns cease, and the timing and speed at which governments seek to replenish reserves. These aspects have the potential to turn what was a counter-cyclical buffer into a pro-cyclical one. The observed price fall that followed the MOU is exactly the signal that brings suppressed demand back to the spot market. David Fishman of the Lantau Group set it out explicitly: “The thing that can’t be sustained forever is the stockpiles of crude. If prices weakened, you’d expect the first thing they do is start to stockpile again.” Returning demand will hit the market at the same moment the IEA’s coordinated 400 million barrel emergency release is set to taper to zero by October, and commercial storage, including at Cushing, is already drawing toward operational floors last tested in 1983.

The bulk of the 2026 US SPR releases have been structured as exchanges rather than sales – barrels out immediately, barrels back under contract – so Washington is not an immediate net open-market buyer. That said, Trump has stated he retains the option of “resetting the clock” if he believes Iran is “not complying”, which would require immediate reserve rebuilding. However, the 60-day window is too short to restore that buffer to a level that would provide meaningful leverage, making the reset option less available than it sounds, and suggesting a binary outcome – either a rapid rebuild or a delayed one, with dramatically different price implications. The same is not true elsewhere, though. Japan, South Korea, and Europe contributed substantial volumes to the IEA release from industry obligation stocks and face a mandatory open-market rebuild.

Suppressed Chinese demand has eased the shock during the closure but now compounds the recovery problem. Societe Generale analysts described China as “the invisible hand that is rebalancing the market,” estimating it curbed oil imports by around 3 mb/d – an amount nearly equal to Japan’s total crude demand – by drawing down onshore stockpiles and limiting refined product exports rather than competing for scarce seaborne barrels. China’s centrality to the crude price now creates a modelling problem that was not foreseen. The traditional price variables – US SPR levels, EIA inventory data, Cushing stocks, rig counts – were designed for a market where the US was the marginal consumer and the marginal reserve manager. China has taken both roles, through import volumes that can swing by 3 mb/d and massive combined strategic and commercial reserves estimated at 1.2 billion barrels.

Compounding the forecasting challenge is the distinction between the operational floor and the political floor. The operational floor is the engineering minimum below which tank farms and pipelines cease to function. The political floor is the level at which a government intervenes to stop further draws, and it lies above the operational floor by a margin that varies with political circumstances but is never published. For the US, Trump’s admission at the G7 that reserves would run out in “about four weeks” was a rare public signal of where the political floor lies.

Everything about China’s strategic reserve is opaque – size, operational floor, political floor – as its management is a sovereign policy instrument rather than a reportable obligation. Unlike the US SPR, it has no exchange mechanism, so its rebuild will be a sovereign purchasing decision. It will have an internal threshold – the political floor – below which stockpiles cannot fall, so China’s re-entry into the spot market is driven by that threshold, and the timing cannot be inferred from price data alone. Yet it also has an established pattern of aggressive accumulation whenever prices fall and many of the tankers now leaving the Gulf are carrying Iranian crude made available by the sanctions waiver, and most appear headed for China – the buyer that has absorbed the bulk of sanctioned Iranian oil and has a strong incentive to restock at current prices. Hence, the market is pricing a recovery whose two largest demand variables – the scale and timing of Chinese re-accumulation – are determined by political decisions that are invisible from outside.

Supply glut?

Despite the turmoil the IEA has returned to its oil glut prediction. Its June OMR forecast a 4.7 mb/d oversupply in 2027 and this deserves the same scrutiny as its January 2026 forecast of a comfortable 2026 surplus. That rested on undisturbed supply and was overtaken by events within eight weeks of publication. The June projection rests on an equally optimistic restoration assumption. The IEA’s own May OMR said rebuilding the cumulative 900 mb stock deficit would require an extra 1 mb/d of supply for three years on top of underlying demand growth; the EIA’s June STEO assumes flows resume in Q3 but full trade-pattern restoration not before early 2027, in line with ADNOC and the IEA’s own assumptions. There will be a surplus eventually, but the question is whether it occurs before or after the demand recovery and reserve rebuild push prices higher in the fourth quarter. The sequencing outlined above suggests later rather than sooner.

The wave

The MOU does lower the ceiling of what the wave might have reached. The market had consistently refused to price the worst case – a Strait that stays shut well into the second half of the year – and the MOU removes that tail risk, which is why the war premium came off correctly. But the market has gone further, pricing as though the supply shortfall itself has cleared, and it has not. The floor is set by the physical reconstitution lag, not by the diplomatic calendar. The wave arrives smaller than it would have under continued escalation, because the ceiling is lower; it arrives larger than the current price allows, because the floor has not moved.

Brent managed money short positions have reached an all-time record, according to Goldman Sachs data – above $16 billion as of mid-June, well beyond the prior peaks of 2022 and 2024. When the physical data establishes what the financial positioning has denied, short-covering will become a second amplifier, accelerating the upward price move.

Crack spreads – the refinery margin between crude input and refined product output – have widened sharply into the falling crude price, with RBOB gasoline futures rising as WTI has fallen since the MOU. That divergence illustrates the two-track argument in real time: the crude price is being pushed lower by financial positioning while physical product markets are experiencing supply shortfalls. A market that has correctly priced a supply recovery produces stable or narrowing crack spreads, not widening ones. The spreads confirm that downstream physical tightness persists even as the headline crude price suggests it does not.

Whether the deal survives

The viability and longevity of the deal itself remains unsettled – the pattern of the first week echoing the observed reality of recent months.

Iran announced a fresh Hormuz closure on 20 June in response to continued Israeli strikes in Lebanon, and the Iranian delegation appeared to pull back from the Switzerland talks the following day, but the talks proceeded, concluding on 22 June with Iran’s Foreign Minister describing “major progress”. The talks also produced a communications line to manage Hormuz incidents. Iran and Oman subsequently formed a joint committee on Strait administration – an institution that reinforces Ghalibaf’s claim, in Iranian state-media remarks on his flight from Switzerland, that the Strait would never return to its pre-war situation. If that is even partly correct, the deal has not resolved the Strait question so much as deferred the dispute over who controls practical access to it.

The US Treasury issued a 60-day sanctions licence allowing the production, delivery, and sale of Iranian oil at market prices – a genuine concession clearing the path for Iranian crude to return to market at full price. Ghalibaf confirmed an agreement to release $12 billion in frozen assets, though the conditions remain disputed: Vance said the funds would require Qatar and US oversight and be restricted to US agricultural purchases; Iran’s central bank governor said Iran had no such obligation. Qatar and Pakistan confirmed a High Level Committee, four working groups, a communications line for Hormuz, and a de-confliction cell for Lebanon – the last of which Pezeshkian presented as a victory, saying the US had “been compelled to retreat on the issue of Lebanon”. Notably, the MOU contains no requirement for Iran to limit its missile programme; Trump stated at the G7 that “missiles aren’t the problem”, dropping one of the war’s many stated justifications.

The nuclear inspectors question shows how far apart the two sides remain. Vance claimed Iran had agreed to invite IAEA inspectors back, but Iran’s foreign ministry said it had made “no new commitments”. Trump then posted that Iran had agreed to inspections “long into the future”, adding that without this concession “there would be no further negotiations” – a public ultimatum delivered while technical talks were still under way. Vance led the Switzerland talks while Trump simultaneously threatened renewed strikes if Hezbollah kept attacking Israel, warned Iran’s leaders they “won’t even make it back” to their country if Hormuz closed again, and proposed imposing US tolls on the Strait after the 60-day period. Ian Bremmer, president of Eurasia Group, has argued on X that Vance says nothing on Iran or Israel that Trump does not want said – which would make the contradiction deliberate rather than chaotic, a good-cop/bad-cop posture that preserves Trump’s option to collapse the deal while Vance builds it. Whether Bremmer’s interpretation is correct or not, Iran is negotiating with a counterparty whose principals are saying opposite things in public, and the deal’s durability depends on which of them is actually setting policy.

The Lebanon ceasefire has largely held since the talks, though Israel has explicitly reserved the right to continue operations and has not withdrawn from southern Lebanon. Iran’s blocking condition – that other provisions of the memorandum will not enter the implementation phase until the war in Lebanon ends – remains formally on the record, and has not been met.

Outlook uncertain on magnitude, not direction

The 60-day period covers three partially overlapping things – the negotiating window for a final agreement, the Treasury sanctions licence, and Trump’s stated reset option if the deal collapses – that face the same precondition Washington may not be able to deliver, an end to the Lebanon conflict. A collapse of the deal would put the war premium back on top of a deficit premium the market has already mispriced. A reminder that the original JCPOA took nearly two years of intense negotiations and there there are fewer than 60 days left in this window. This deal is really another ceasefire extension, with all the uncertainty that entails both during the window and after it closes.

Upward pressure holds regardless of the diplomatic calendar, because of the gap between paper and physical reality. A further price rise into the third or fourth quarter does not need Iran to renege or Lebanon to reignite the wider war – it needs only the lag to hold, and the shipping and production data confirm that it should. Whether or not the deal survives, demand recovery will arrive before supply recovery, with storage buffers already drawn close to their floors, and all buyers competing in the same constrained market at the same time. Mohamed El-Erian, President of Queens’ College, Cambridge, and former CEO of PIMCO, writing days after the signing, set out the operational complexity – Hormuz does not reopen at the flick of a switch, given the insurance, demining, and production ramp involved – but his caution stays qualitative where the constraint is quantitative and timed.

The Islamabad MOU removed the war premium from oil while leaving the deficit premium untouched. The deal changes the size of the wave still to come, not whether it arrives.

The remaining five channels – the Gulf’s confidence economy, three simultaneous capital withdrawals from the US asset ecosystem, the US security commitment to Taiwan, Europe’s rearmament against a guarantee that is retrenching, and the AI capital expenditure cycle – will be addressed in follow-up pieces.

Apologies for the delay in publishing. The scene has been moving across multiple theatres simultaneously, and calibrating it has taken longer than normal.

President Trump and Iran’s President Pezeshkian formally signed the memorandum of understanding (MOU) on 17 June, after Vice President JD Vance and Iran’s parliamentary speaker Mohammad Baqer Qalibaf electronically signed the MOU on 15 June, in both English and Persian at Tehran’s insistence. Trump signed his copy at the Palace of Versailles before a dinner hosted by French President Macron. “It’s signed,” he told reporters afterward, “Signed in Versailles. Just signed it.” Iran’s Foreign Ministry has called the document already in effect, though an MOU is not a binding treaty, requires no Senate ratification, and can be abandoned by either side without formal consequence. Trump has reserved the right to resume bombing if the terms are not met. The formal ceremony once scheduled for Geneva on the 19th has now been cancelled, but negotiating teams from both sides are expected to travel there for working talks.

The MOU is closer to another ceasefire extension, this time with a fixed 60-day period, than to the unconditional surrender Trump promised in March. Iran has had its supreme leader killed and major nuclear sites struck, and thousands of Iranians have died getting to this point, but the regime has survived and is still holding most of what its ‘survive and exhaust’ doctrine set out to achieve, regime survival paired with coercive leverage over a chokepoint, not a conventional military win and certainly not an improvement in ordinary Iranians’ lives.

Looking at Trump’s own varying objectives, the war’s results do not amount to much. Ending a conflict Washington itself started is not a strategic gain. Reopening a waterway that was open before the war, on terms that may yet let Iran charge for the privilege, is not a strategic gain. And preventing Iran from ever acquiring a nuclear weapon remains exactly as unresolved as it was pre-war, if not less resolved than it was under the JCPOA deal Trump abandoned in 2018. Senator Bill Cassidy put the domestic verdict bluntly: ‘before the war, the strait was open and Iran was being crushed by sanctions; now thirteen Americans are dead, families have paid billions at the pump, and sanctions are set to be lifted anyway’.

As an asymmetric conflict, conventional measures of victory are misleading, and by unconventional measures this has been an Iranian victory. The 14 paragraphs of the released text, read aloud to reporters by a senior US official and annotated by the New York Times paragraph by paragraph show why.

The false premises, paragraphs 2, 8 and 9

Judged by conventional metrics in March, Iran was losing. Ballistic missile launches were down sharply, nuclear infrastructure had been struck, and Khamenei himself was dead. Yet, judged against Iran’s own strategic objectives, the picture looked different even then. The regime had demonstrated it could not be destroyed from the air, while it could make continued American involvement politically unsustainable, split Washington from its Gulf partners, and ensure that any successor leadership emerged hardened rather than chastened. ‘Survive and exhaust’ is the recent term for the doctrine, but the strategy traces back to lessons the Iranian Revolutionary Guard Corps (IRGC) drew from the Iran-Iraq War four decades ago. A regime insider speaking to the Financial Times this week put the same logic more simply: the war fitted perfectly into the ideology Tehran had spent decades preparing for, and it strengthened the system rather than breaking it.

That doctrine worked because the premises for fighting the war did not hold up, and the war was launched, as I have argued before, without a settled definition of victory. Trump declared the war won “in the first hour”, then said he would know it was over when he “felt it in my bones”. The stated objectives shifted accordingly: dismantling the nuclear programme, preventing enrichment, destroying the missile arsenal, freeing the Iranian people, and regime change, each promoted and then quietly dropped in turn.

The risk of Iran closing Hormuz was a consistent warning that was ignored, not an intelligence failure. The Wall Street Journal reported that Joint Chiefs Chairman General Dan Caine warned Trump in multiple pre-strike briefings that US officials had long believed Iran would deploy mines, drones, and missiles to close the strait if attacked. CNN confirmed separately that the inter-agency analysis from Treasury, Energy, and State – the kind of assessment integral to planning in past administrations – was treated as a secondary consideration rather than a primary input. Every administration from Carter’s through Biden’s had taken the strait risk seriously enough to avoid direct military confrontation with Iran on Iranian soil, not because they lacked the capability, but because the wargames told them what the economics would look like. The Trump administration had those same wargames, yet it chose to proceed anyway.

Paragraph 8 of the released text shows where the central premise regarding Iran’s nuclear enrichment actually stood. Iran held around 9.9 tonnes of enriched uranium, including roughly 433 kilograms of uranium in UF6 enriched up to 60%, as verified by the IAEA’s February assessment. This is a level no other non-nuclear-weapon state inside the NPT has reached. That was well beyond ordinary civilian activity and into dangerous threshold status, a genuine proliferation concern, not innocent behaviour, and the IAEA’s own warning that it could no longer maintain continuity of knowledge over the stockpile after losing access only heightened that concern. But threshold status is not the same as an active weapons sprint.

A state genuinely racing to a bomb has every reason to take the short remaining step to weapons grade, leave the NPT, and present the world with a fait accompli, the way North Korea did, withdrawing from the treaty in January 2003 and testing its first device in October 2006. By every indication available before the IAEA lost its access, Iran had done none of that: it had not verifiably crossed beyond 60%, withdrawn from the NPT, expelled inspectors altogether, or prepared a test. Paragraph 8 reaffirms that Iran will not procure or develop nuclear weapons, which is not a new concession; it dates to Iran’s 1970 ratification of the NPT and was repeated upfront in the JCPOA. The roughly ten tonnes of enriched material Iran holds is to be down-blended on site under IAEA supervision, but it does not have to leave the country, which is a step back from 2015, when Iran shipped 97% of its stockpile to Russia. Whether Iran keeps enriching at all, and for how long any new constraint runs, is left entirely to the negotiation set at 60 days by paragraph 3. Paragraph 9 commits both sides to a status quo in the meantime: Iran holds its nuclear programme exactly where the strikes left it, and Washington imposes no new sanctions and deploys no further forces. The question this war was ostensibly fought to answer remains just as open as it was on 27 February, frozen rather than resolved.

Paragraph 2’s mutual pledge of non-interference discreetly buries a different justification that floated around the war’s opening weeks: that this was, in some sense, also a humanitarian intervention, freeing the Iranian people from their own government. That framing borrowed the language of the Responsibility to Protect (R2P) doctrine without meeting any of its conditions. R2P was created to protect populations facing mass atrocity, regardless of nationality, through multilateral authorisation and a stated protection mandate, not through a unilateral bombing campaign aimed at military and nuclear targets. The US has no record of invoking it for any peoples on a timely basis or for Iranians specifically: not during the Green Movement in 2009, not in the 2019 or 2022 protests, and not during the killings of January’s uprising. Paragraph 2 puts the honest and accurate position in writing – Iran’s domestic political future is now explicitly Washington’s business to ignore.

The deterrent Iran already had, paragraphs 4 and 5

The threat of closing Hormuz did not deter the attack that started this war, but it did make continuing the war, once started, strategically unsustainable for Washington, and gave Iran a way to force an off-ramp that did not depend on winning a single battle. Iran did not need a nuclear weapon to achieve that. The nuclear programme was serving a separate purpose, prestige and international standing, with elements of bargaining leverage and deliberate latency mixed in, rather than a straightforward sprint toward a bomb. A demonstration that a state can do something technologically demanding without foreign permission is valuable for the standing it confers, in a manner not unlike the Apollo programme for the US. Sina Azodi’s recent book, Iran and the Bomb, reaches a closely related conclusion from the historical record rather than from this war’s events. Both the Shah’s government and the Islamic Republic built nuclear hedging, the academic term for deliberate latency short of weaponisation, for reasons of prestige and standing as much as military necessity. Iran’s own foreign minister has described the programme in exactly those terms, as a matter of national pride and dignity.

The success of the US-Israeli air campaign paradoxically ended up working against those two belligerents. Before the war, Iran retained enough of its asymmetric maritime toolkit, mines, missiles, fast boats, coastal systems, and a demonstrated willingness to use them, to keep the Hormuz threat credible without needing its full conventional military intact. Even so, deploying that toolkit in full risked drawing further strikes on conventional assets Iran still had reason to protect. Yet, once the air campaign had degraded those assets, that residual restraint largely disappeared. So the war did not just fail to suppress the asymmetric threat it was meant to forestall, it actively unlocked it. By attacking a country whose principal retaliatory lever was the world’s most important oil chokepoint, the campaign handed Iran the occasion to demonstrate that lever’s worth in practice rather than merely in theory. Iranian state media has not been shy about saying so. Tasnim, a news agency closely aligned with the IRGC, declared that no outside power can define the security order in West Asia without accounting for Iran’s role and power. Ray Takeyh of the Council on Foreign Relations put the underlying point about the strait more pointedly still: a fee for transit is one form of monetising the geography, but payment for not attacking the Gulf states is another, and the second is the prerogative of a tributary state rather than a port authority. There is a reason, he notes drily, it is called the Persian Gulf.

Paragraph 4 has the US beginning to lift the naval blockade immediately on signing and ending it fully within 30 days, restoring Iranian export flow, overwhelmingly bound for China, well before the harder nuclear negotiation even starts. The New York Times’s David Sanger correctly interprets this as Washington giving up one of its few sources of leverage before the negotiation that leverage was meant to support has begun, hence giving Iran every incentive to let 60 days stretch into months. Paragraph 5 carries the same unresolved edge for the strait itself. Free passage applies for 60 days only. Nothing in the text stops Iran trying to charge fees afterward that it never charged before the war, though a unilateral transit toll through an international strait is legally contested under UNCLOS which does not allow coastal states to demand payment merely for passage, even if narrower fees for specific services may be defensible. Whatever Iran can lawfully charge, the institution to administer it already exists, the Persian Gulf Strait Authority (PGSA), which is now legislated and printed onto published maps, not simply an informal arrangement that lapses when the fighting stops. Mojtaba Khamenei invoked the strait’s closure as leverage in his first public statement since succeeding his father in March. Paragraph 5 commits Iran to negotiate the strait’s longer-term administration with Oman and other Gulf states, but the authority it would be negotiating away, if it ever agrees to, did not exist before the war and has every institutional reason to outlast a 60-day ceasefire.

Two kinds of exhaustion

Decapitation removed Khamenei and senior commanders, but decades of anticipating that scenario meant authority was already decentralised and succession already planned. The regime behaves less like a single command structure than a system, a hydra even: cut off one head and the layers beneath it, the IRGC, the clerical establishment, the succession mechanism itself, simply produce another. What that succession actually produced was not a more compliant regime but a more hardline one. The IRGC, built from its founding to function under external attack, was always the institution best positioned to benefit from a campaign that hit the more exposed clerical and political leadership around it, and the commanders who emerged are younger, more ideologically hardened, and shaped by what they regard as victories in Iraq, Lebanon, and Syria rather than by the caution of a generation scarred by the Iran-Iraq War. Washington achieved a kind of regime change, but it was not the kind it intended.

Washington was exhausted too, and not only politically. The two blockades that ran simultaneously through the war were never symmetric. Iran’s required no sustained operation, just demonstrated willingness backed by mines, fast boats, and coastal missiles built specifically over decades for this contingency. The US blockade demanded daily enforcement across a contested waterway, against a coastline a former NATO Supreme Allied Commander has called impossible to seal completely, with dozens of ships bypassing the line despite more than a dozen warships enforcing it. I have argued from early in the war that magazine depletion, not battlefield outcomes, halts prolonged exchanges between modern militaries and the states absorbing them. Defence analyst Harry Kazianis has estimated that roughly half of America’s THAAD and Patriot interceptor stocks were expended in seven weeks of war. Meanwhile, the Strategic Petroleum Reserve has been drawn down to its lowest level since 1983, depleted by the releases used to suppress crude oil prices during the closure. Trump essentially confirmed the exhaustion constraint at the G7, telling reporters that “we run out of reserves at about four weeks”. Signing the deal as it was set out thus looked less like a concession to Iran than a recognition that the war could not be sustained. The result at home is a deal that has united Trump’s hawkish allies and his anti-war base against him for opposite reasons: the hawks think it gives Tehran too much, the ‘America First’ wing never wanted the war that produced it.

What the war has cost, and who is paying for it

Closing the world’s most important oil chokepoint produced the largest supply disruption in the history of the global oil market – this was the IEA’s own technical assessment, not merely rhetorical flourish. It was deemed larger than the 1973 embargo and Iraq’s invasion of Kuwait, and combining an oil shock with a simultaneous LNG shock was an outcome no prior crisis had produced. Brent moved from $65 before the strikes to above $120 at points during the war. American gasoline prices rose by roughly 38% and diesel by close to 50%. However, from where the market is positioned now it does not look like an ongoing crisis – Brent has fallen back to around $77 on expectations that Hormuz traffic resumes quickly, and headlines suggest that the worst has passed. However, the counterargument is that the worst has been deferred rather than resolved. Mine-clearing alone is expected to take six weeks or more before most shipowners consider the passage safe, and the tankers that migrated to Atlantic routes will not return quickly once it is. Overall Gulf output is expected to reach only 60-70% of pre-war levels by mid-September at best. Meanwhile, the buffers that absorbed the shock – strategic reserves, commercial inventories, suppressed Chinese imports – continue running down on a schedule that does not stop because a memorandum gets signed. The supply side impact of this war is not finished just because the price is subdued.

Iran’s own costs were heavier still, and the regime’s victory claim does not erase them, though the blockade imposed those costs on the wrong target to achieve what it was meant to achieve. Oil export receipts fund Iran’s import basket, not the rial wages that pay the IRGC, so cutting them squeezes households and import-dependent sectors well before it threatens the state’s own operational capacity, a distinction the blockade’s proponents consistently elided. US and Israeli strikes destroyed crucial infrastructure, killed thousands of Iranians, and killed Khamenei himself along with several senior military commanders. Within the establishment, some officials remain reluctant to declare outright victory given the negotiating and economic challenges still ahead, and the test, as one regime figure close to the process put it, is whether the achievements claimed now get converted into lasting gains rather than remaining a wartime narrative. Ordinary Iranians describe a different experience entirely. “The Islamic republic can claim victory”, one Tehran resident told the Financial Times this week, “but many of us do not share that feeling. We feel exhausted, frozen, and abandoned by everyone.” Regime survival and national wellbeing are not the same scorecard, and this piece is only making a claim about the first.

Americans have noticed their own costs, and the polling has been consistent rather than just a single bad month. The recent Reuters/Ipsos survey published, fielded 3 to 8 June, put Trump’s overall approval at 35%, his handling of Iran specifically at 29%, and found 53% of Americans saying the war had not been worth it against 25% who thought it had. Congressional Republicans, including a longtime Iran hawk, responded to the announcement by saying they did not yet know whether the deal was good policy; not opposition, but an absence of the confidence a genuine victory would have produced.

Why 60 days will most likely become another 60 days, paragraphs 3, 12, and 13

Paragraph 3 sets the negotiation at a maximum of 60 days, extendable by mutual consent, and paragraph 13 makes the sequencing explicit rather than just implied: talks on everything else begin only once paragraphs 1, 4, 5, 10, and 11 are already being implemented. Paragraph 12 is the only enforcement mechanism in the entire document, an executive body to monitor compliance, with no detail yet on how it would actually compel either side to act. A structure with a short deadline, a dependency on prior steps before talks even properly start, and no real enforcement behind it is not built to produce a fast or durable outcome.

I have argued previously that the ceasefire process would not hold, and the pattern so far supports that view. An April ceasefire collapsed within days, a mid-April reopening of the strait turned out to be largely fictional, and a full return to US airstrikes followed in early June, just before this MOU was announced. The Gaza ceasefire offers an instructive lesson. Its first phase, the ceasefire and hostage exchange, took effect on 10 October 2025 and was largely completed within days. Its second phase, the harder commitments around Hamas’s disarmament and Israeli withdrawal, was not announced until 14 January 2026, three months later, and has stalled since, with the Board of Peace’s own envoy admitting the deadlock has paralysed everything else. There is no obvious reason the Iran MOU’s harder second phase, enrichment limits and a sanctions schedule, should prove any easier and faster than disarmament has. Takeyh’s own conclusion is much the same: the MOU is best understood as mostly a ceasefire, the kind of arrangement that can be extended indefinitely while a grand bargain is negotiated, or never negotiated at all. The more likely failure mode is not a collapse back into open war but a 60-day extension followed by another, a frozen hot war, for as long as extending serves both sides better than the alternative.

Paragraph 14 commits any eventual final deal to endorsement by a binding UN Security Council resolution, in the manner of Trump’s earlier Gaza agreement. Given how far apart Washington and Tehran remain on the nuclear issue alone, that day may not arrive at all.

The money, paragraphs 6, 7, 10, and 11

Paragraph 7 commits the US to a schedule for terminating, or securing the termination of, UN Security Council resolutions, IAEA Board of Governors resolutions, and unilateral US sanctions. Washington can bind itself but cannot unilaterally end measures other bodies imposed. Paragraph 10 grants Iran immediate Treasury waivers to export crude oil and petroleum products, alongside the banking, insurance, and transport services that exporting actually requires. Paragraph 11 has the US undertake to make Iran’s frozen funds, $24 billion or more, fully available, though the procedures for releasing them are still to be mutually agreed, so it is immediate relief in principle rather than immediate cash in hand. Paragraph 6 commits the US, with regional partners, to a $300 billion reconstruction and development plan, though Trump has said Washington will not put money into it directly, leaving the financing to Gulf states of which he can neither enforce nor prevent contribution. None of these figures has an enforcement mechanism attached yet, so all of them are promises rather than guarantees.

Lebanon, paragraph 1, and the cost to Israel specifically

Paragraph 1 declares the immediate and permanent end of fighting on all fronts, including Lebanon, but it is not a self-executing ceasefire. Israel’s National Security Minister has stated that the deal does not bind Israel, and Hezbollah has consistently rejected the ceasefire terms as binding on its own operations, refusing to accept a deal reached in negotiations it was not party to. What it is, more precisely, is a diplomatic instrument: language Iran can use to pressure Washington over Israeli conduct in Lebanon, regardless of whether Israel actually changes that conduct. Iranian foreign minister Abbas Araghchi told assembled diplomats this week that Iran considers the MOU’s parties to be the US and Israel on one side and Iran and Hezbollah on the other, and that without Israeli withdrawal from occupied Lebanese territory, the war cannot be considered fully concluded.

Netanyahu now holds Israel’s security zone in southern Lebanon at real political cost, risking a breach with Washington that domestic opponents are already using against him, because Israel’s war aims were never the same as America’s. Israel sought the Iranian regime’s collapse and Hezbollah’s destruction, a far more absolute goal than ending an enrichment programme. I asked in March what Netanyahu would be left with once the US engagement he had pursued for a decade came to an end. The answer is now visible: an agreement Iran can plausibly interpret as siding with its own position on Lebanon, an IDF chief of staff who had already warned the force could not sustain the operational tempo needed to consolidate the campaign’s gains, and a domestic position weak enough that defying the deal is itself a political risk rather than a free option.

Real strain inside Iran, but it is not the regime cracking

The IRGC suppressed Iran’s own January 2026 uprising before the war began, and Pezeshkian’s presidency has since accused senior IRGC commanders of escalating the conflict unilaterally, against the government’s own negotiating position. Hardline figures have reportedly chanted for the deaths of Araghchi and Qalibaf over the agreement. Although that looks like a regime under threat, it is more accurately a regime arguing with itself over method, since the people doing the chanting are not calling for the Islamic Republic’s end, but calling its negotiators traitors to a system both sides still want to preserve. The executions continuing since January, tied to the unrest and to wartime espionage accusations, reinforce that conclusion: a state demonstrating coercive control and consolidating it is different from one losing its grip, even if the two look similar from outside.

One hundred days after assuming power, no photograph of Mojtaba Khamenei has emerged as proof of life, and several senior IRGC generals remain underground. In a system where visible authority signals institutional continuity, those absences suggest the succession remains contested or incomplete rather than settled. Factionalism was always how Iran’s system operated, managed for decades by a supreme leader whose job was to arbitrate between competing power centres rather than eliminate them. The war removed the arbitrator, however, and so the factions it left behind are still arguing inside the same system, not trying to bring it down. Saeed Jalili, the hardliner standard-bearer, has reportedly been removed from the Supreme National Security Council, while Ghalibaf secured re-election as parliamentary speaker with more than 80% of votes, suggesting the pragmatist faction is consolidating its position even as the hardliners remain vocal. The hardliners’ specific fear, articulated openly at rallies, is that Trump will strike again once Hormuz is open and global oil reserves are replenished, having used the deal merely to resupply.

What the demonstration costs Iran later

Iran’s win has a price attached, though it is a price with a long fuse. Having shown that Hormuz could be closed and held closed for nearly four months, Iran has also shown every major energy importer, Gulf neighbour, insurer, and tanker owner exactly how exposed they are to a single actor’s control of a chokepoint nobody can route around quickly. Effective alternatives to that exposure – new pipelines bypassing the strait, long-term contracts diversified away from Gulf crude, and naval arrangements that do not depend on the US Navy alone – take years to build and finance, not months. For as long as those alternatives do not exist, which on any realistic construction timeline is most of the next several years, Iran’s leverage over the strait survives this war fully intact, regardless of how the politics of this MOU resolve. The Houthi precedent across the Bab el-Mandeb is a reminder that Hormuz is not even Iran’s only card of this kind. The leverage demonstrated this year will eventually be worth less once the world has had time to build around it, but that erosion has not started yet.

Iran has secured, by the doctrine it set out to prove from the war’s first weeks, most of what ‘survive and exhaust’ was designed to deliver: a surviving regime, a politically exhausted opponent, a fractured American relationship with the Gulf, and a hardened successor leadership that believes, with some justification, that it won. Trump promised unconditional surrender. Instead, he accepted a ceasefire document, a 60-day clock that has already been suspended once, and an Iran that controls the world’s most important oil chokepoint on terms it wrote. By the only scorecard that was ever going to matter in an asymmetric conflict, Iran is ahead.

Postscript

Update, 19 June: Within 24 hours of the formal signing at Versailles, Iran suspended its negotiating team’s planned departure to Switzerland, citing ongoing Israeli strikes in Lebanon as a breach of paragraph 1. Araghchi stated that any Israeli military action in Lebanon constitutes a violation of the MOU across all fronts. Trump responded on Truth Social that talks were “continuing at a rapid pace”. Iran said they were not. In March, The Deal That Keeps Not Happening argued that every framework fails at the same junction, when the party with a structural interest in ensuring no deal survives finds its trigger point. Lebanon was always that trigger point and it took less than 24 hours.