The floor is demonstrated; the re-arm is capped.

Warehouse stocks broke below 300kt for the first time since 2022 — 287,725t by 13 July, near 20 year lows — and the LME turned on it, rebounding from a 4 month low of $3,085 to ~$3,155 as physical scarcity fired rather than the war returning. The premium the 8 July ceasefire collapse should have re-armed onto that thinned base has stayed capped: EGA's Al Taweelah is restarting ahead of schedule and Chinese and Indonesian output keeps climbing, so the metal held rangebound rather than spiking, the December 2027 contract still at a discount. The genuine tightening sits beneath the metal, in the cost stack — alumina firmed to ~$331/t against a capped LME, and Gulf gas impaired at both ends is lifting smelter power costs, squeezing survivor margins from below. Downstream, Airbus delivered its best H1 since 2019; with engine bottlenecks easing, the Gulf billet outage is climbing the constraint queue towards the 2027 rate ramp.

The war is drag, not the cap, and the real tests land in late July.

June looked like strength — a 16.52M SAAR, up 4.4% YoY for a second straight monthly gain even as pump relief vanished with Brent back above $85 — but it was bought, not earned: incentives ran $3,217 a unit (+12.7%), the average monthly payment hit a record $813, and retail fell 4.1% across the first half. The ceiling held because OEMs discounted into it. What actually caps the sector sits outside the war and converges in late July: the tariff cliff is days out — the Section 301 successor completes and USMCA Round 3 opens on 20 July, the Section 122 surcharge sunsets on 24 July — and it's a hand-off, not a reprieve, 301 backfilling 122 at no cap and no expiry after Washington declined to renew USMCA on 1 July. Underneath, the mix keeps tilting to hybrids (+19.4% H1) over BEVs (−25.1%), and the China to Japan magnet axis stays truce proof — dysprosium and terbium oxide to Japan were halted with no clock, exports there off 34.6% m/m in May. The real reveal is still ahead: Q2 earnings strip the one time IEEPA refund and any easing input cushion, landing straight into the tariff week.

Record fuel, but the margin held — and the corridor hardened.

Both halves of the sector turned adverse this quarter, for different reasons. Fuel hit a record: Delta absorbed the highest quarterly fuel expense in its history, up 77%, United's up 84%. Yet neither margin collapsed — both beat and held or raised guidance by passing the cost straight through into inelastic demand, fares up ~20% with no visible elasticity. Delta posted record revenue, reaffirmed its outlook for the year and lifted the dividend 15%, recapturing about 60% of the fuel increase; United expects the rest back by Q4. Jet fuel has since eased ~20% off its Q2 peak but stays volatile, and the floor beneath it is indifferent to the war — US jet supply is the tightest in over 60 years, days of supply near 21, a refining constraint no ceasefire can move. The corridor is the half that hardened: EASA re-issued a bulletin on 14 July advising operators to avoid the Bahrain, Kuwait, Qatar and UAE airspace, Kuwait's overflight ban runs to 4 August, and a fresh wave of strikes disrupted more than 1,500 flights. The Gulf carriers route around it at 75–96% of operations and take share from the rivals suspending into the risk, while the hedged and integrated defend the quarter and the most exposed suspend guidance or collapse.

The squeeze is real but driven by cost, and it lands on the wrong chemistry.

The margin scissor everyone read off the nickel tape needs re-reading. The cost side did re-arm — the war re-strands Gulf sulphur, which runs 30–35% of HPAL operating cost, on top of China's acid ban and Russia's damaged capacity, and ore grades below 1.5% burn more acid per tonne of nickel, so the pressure is structural rather than a passing spike. But the price collapse was overstated: refined nickel's ~14% June fall ran on a rumour of quota loosening that Indonesia then denied on 24 June, and the price HPAL actually realises — battery grade nickel sulphate — held firm even as the LME headline dropped, because tight quotas pushed integrated smelters towards stainless steel. The squeeze is led by cost, and shallower than the tape implies. Sharper still, it lands on the wrong chemistry: the constrained nickel feeds NMC and NCA, where EV demand is soft, while the booming grid storage on the AI build runs almost entirely on LFP, which needs no nickel or cobalt. That surge's binding constraint isn't metal — it's the transformers and switchgear the integrated storage blocks now bundle, the same bottleneck throttling the data centres themselves. The squeezed input feeds the soft segment; the growing one sidesteps it entirely.

The bottleneck is the story, not the war.

The binding constraint here sits entirely outside the war, and it hardened this cycle on its own clock. Hyperscaler 2026 capex firmed towards ~$750bn against a grid that can energise only about a third of announced US capacity: high voltage transformer lead times run 128–144 weeks, up to 4 or 5 years for the worst custom units, and a campus joining the Northern Virginia queue now cannot realistically draw power before the early 2030s. Electrical gear is under 10% of a data centre's cost but 100% of its critical path, so capital cannot outspend the bottleneck — and the biggest buyers stopped waiting. Power announced behind the meter in 2025 alone reached ~50 GW: Oracle islanding Project Jupiter, Meta running Hyperion on gas, xAI trucking generators into Memphis. Power delivery, not compute or capital, is what binds. Beneath it the Section 232 stack strikes the transformer's own copper windings, leaving the most inflated construction input stack on record; the war reaches only the minor offsets, as diesel, feedstock and copper re-firm to their highest in 3 weeks. On demand, the homebuilder reset is a hard print — Lennar's gross margin down to 15.6% from 17.8% — while a bill on housing affordability cleared Congress only to have its signing abruptly cancelled and rates head higher.

The consumer split in two: spending resilient, sentiment near stress.

The mechanism that runs pump prices into sentiment is working forward again. Petrol re-rose through early July, snapping 8 straight weeks of declines, and — the telling part — held there even as crude softened mid month, confirming the refining floor that keeps the pump elevated on tight product supply rather than the crude the daily headlines quote. That cuts the ground from under June's sentiment reading, which had inflected up off May's record low on the specific expectation that the conflict damage was temporary — a judgement the resumed strikes refute. The sharper development, though, is a split in the data itself. The hard numbers held: June spending stayed resilient, retail a modest gain, Q2 consumption tracking around 2% annualised against Q1's 0.5%, carried by tax refunds, firm equities and a steady labour market. The soft numbers didn't; sentiment sits near stress. The question is not whether sentiment weakens but whether it drags the resilient spending down with it. June's CPI landed soft, but the print caught the energy low of the reopening before the pump turned — a lagging snapshot, with the re-rise registering in July. Beneath it, inflation expectations are re-arming into a fresh energy shock, so a re-weakening consumer meets a central bank with less room to ease, not more.

The stress left the barrel for the still.

The war premium is climbing again — Brent back above $85 on the resumed strikes and the reimposed closure — and any softer number on the screen is a fiction, priced off a reopening that never physically happened. Underneath the price, though, the physical barrel loosened: the run of 10 weekly US draws broke, Cushing built off its floor, the strategic reserve draw halved, and global inventories rose in June for the first time in 4 months, as distressed Russian crude and disciplined US supply refilled the balance around the gated Strait. Comfortable barrels, not a comfortable price — and the tightness that matters has moved downstream anyway, to the still. Two refining wars are pulling capacity out of the same short market — Gulf damage now reaching into gas processing, Russian throughput at its lowest in 21 years — and Russia's outright ban on diesel exports has fired, removing roughly 11% of global diesel supply and carrying the European diesel margin and the US 3-2-1 crack to record highs. For every sector downstream the cost signal has changed: no longer the Brent flat price but the diesel crack, the petrochemical outage map and the gas balance of the region you operate in — because gas itself fractured 3 ways, the US insulated on home supply, Europe refilling behind pace toward a thin winter, and Asia paying the premium into its cooling peak. Crude can take a moderate shock; refining, once burned, cannot.

The front end eased, but the back end, the energy bill and the insurance gate held.

The rate trap the sector feared has half inverted. At the front, June's core inflation came in cold — flat on the month, led by the smallest shelter rise since early 2021 — and a July hold is now better than 90% priced, so the funding cost the chain borrows at eased rather than re-armed. Two things keep that from being an all clear. June's headline energy relief was priced off a Strait reopening that never physically materialised, so the real pass through — a renewed crude premium, record refining cracks, airline fares already up 26.5% y/y — is deferred to the July prints, not resolved. And the long bond held on its own terms: it tailed back through 5% on fiscal supply rather than inflation, with foreign officials now net sellers of duration and the replacement bid a private one that runs on leverage — so leveraged borrowers get no relief at the back end from the cold core. Underneath, the insurance gate hardened into a regime of permits and blockade, hull war cover holding wide and the vessel toll now reaching a Gulf state's own tankers. The clean positive is the banks themselves: the institutions that finance trade just posted the best quarter in US banking history — strong enough, so far, to absorb the private credit and AI capex stresses sitting beneath them, though the deferral underneath keeps stacking.

Grain and beef are at the shelf now — the fertiliser re-arm is still upstream.

The nitrogen benchmark has turned back up — CBOT urea about 16% higher on the month as the re-closed Strait pushes diesel and Gulf sulphur costs back into fertiliser logistics — but the shelf hasn't felt it. US retail urea is still falling, down around 13% and easing into July: the retail channel runs weeks behind the benchmark, so what farmers are paying now reflects June's ceasefire low, not the market that replaced it. The gap between the two is the bill still to arrive, and it lands in the second half. The phosphate floor beneath both never moved, sitting at its highest since 2022 on China's export ban and destroyed Gulf sulphur. What the shopper is actually paying for now is elsewhere. The grain buffer's forward failure has gone from forecast to balance sheet: the July WASDE cut US wheat ending stocks to their lowest in years, on the smallest harvested acreage in roughly 150 years, and trimmed world corn on European heat, with a record US soybean crop the only offset. And protein is the live grocery driver — beef up 12% y/y, ground beef at a record on the smallest US herd in decades, eggs and butter the one reprieve. Food inflation reads soft, near 2.7%, but that print caught the trough before the escalation; the pressure is building, not absent.

The acid squeeze has three sources now, not two.

The sulphuric acid chokepoint that runs under batteries, alumina, semiconductors and fertiliser now has three independent sources, not two. The Gulf war keeps Middle East sulphur stranded behind the Strait; China's ban on more than 40% of world acid output holds; and Russia's export ban, driven by war damage and extended after the Astrakhan strikes, is now a confirmed third leg. No single de-escalation relaxes the market — it takes three, and acid landed costs sit at more than double their pre-war level. Copper makes the same argument from the other direction: the early July macro dip was bought back within a fortnight, the metal returning to its highest in 3 weeks through a war escalation, as Chile's falling output and converging deficit forecasts held the physical floor while the paper did the moving. The US tariff decision on refined copper cathode is still pending over a record COMEX stockpile — delivered by Commerce, awaiting the President. Nickel is the divergence worth flagging: its screen price fell on an Indonesian move to loosen quotas and a record exchange overhang, yet the battery grade material the acid squeeze actually bites stays tight — a split between paper and physical inside one metal, the exposed seaborne buyers carrying a cost the integrated producers are shielded from.

This constraint is self imposed, and the first test just passed.

Pharma is the one sector in the set whose binding constraint isn't the war — it's a self imposed US policy, the 100% Section 232 tariff that hits the 17 largest companies at the end of July and everyone else in late September. The war reaches only the input costs a step down the chain, and there it climbed back like everywhere else: the sulphuric acid behind common APIs sits on the same squeeze from 3 sources, petrochemical feedstock and blister foil firmed as the Strait shut again, while medical helium stayed rationed on Ras Laffan's outage of years. The event that matters is that the binding test moved from forecast to first data point. J&J reported first and confirmed the core call: the branded major, dealed and protected on price, named the tariff as a margin factor but guided to recoup it — a pretax margin improvement near 75 basis points — and raised its outlook for the year. That is the shielded tier passing. The unshielded tier is the fault line: generics, carved out of both the tariff and the pricing regime, carry the same re-arming cost with no offset and an exemption that itself expires in 2027. And the shortage count rose — 223 active in Q1 — but stayed structural, sterile injectables and generics failing on quality and margin, not one break tied to the war. Cost, not disruption, held again.

The last flex closed.

Tyres are the only sector in the multiway competition for constrained sulphur with no chemical substitute for what they need it for — vulcanisation has been the rubber to tyre process since 1839, with no alternative at road scale — and they sit at the bottom of the margin hierarchy, so they get outbid. Mining runs on captive acid, batteries pivot chemistry, farmers shift crop mix; tyre makers cannot. What they did have was one flex: substituting natural rubber against synthetic when either ran dear. That flex has now closed. The synthetic side rose with oil, lifting SBR, butadiene and carbon black across roughly 70% of the input stack, while natural rubber climbed on its own account — up double digits year to date on a structural deficit near 400,000 tonnes, as ageing trees past peak yield and weather shortened tapping windows bite across the Southeast Asian base that grows about 80% of world supply. Both elastomers moving up together leaves no gap to substitute into. Only steel cord holds stable. The sector compresses rather than breaks, but on a narrower base than before — premium buffers and pricing power, not a cheap input. The one clean relief is European: definitive anti-dumping duties on Chinese passenger and light truck tyres came into force on 8 July, though the 24.4% cooperating rate catches the Western brands' own Chinese plants, so the benefit favours EU made over China made rather than one cohort over another.

The inputs held and the trade broke; the two aren't the same thing.

The input map passed its test on both directions of the war. Transit bound inputs stayed constrained as the Strait re-closed — Gulf sulphur, roughly 200 helium cryo-containers still frozen under the insurance gate — while the China controlled layer sat exactly where the tech war put it, indifferent to the Strait because none of it moves through the Strait. The tungsten hexafluoride cliff fired on schedule: 2 Japanese producers, together about 30% of world capacity, permanently ceased in July, with contracts locked 70–90% higher. And the concentration bottleneck never moved — advanced packaging remains the ceiling on AI accelerators, sold out through year end, confirmed by TSMC's record June. What broke was the trade, not the physics. SK Hynix completed the largest foreign US listing ever on 10 July, then fell 15.4% in Seoul days later on a single brokerage note about HBM4 shipment timing, tripping a market wide circuit breaker; the memory complex is in a bear market, roughly $1.5tn of semiconductor value gone since late June, and Samsung's record quarter couldn't lift its own stock. That is positioning, not a demand crack. But underneath it sits one real physical signal: contract DRAM price gains for Q3 are guided to 13–18%, down from about 60% in Q2. Still short, still rising, decelerating.

The gate is a permit booth, not a wall — and now it's gated twice.

The near standstill read of early July was wrong, and the correction changes the shape of the risk rather than its direction. The Strait is not physically sealed. Iran administers a declared closure by permit through its own strait authority — a closure it cannot enforce physically, and enforces instead by making the open lane lethal, demonstrated when a container ship was struck on the very southern lane under Omani negotiation. Overlaid since 14 July is a reimposed US naval blockade of Iranian ports, all flags, more than 20 warships. Two rival administrative regimes now govern one waterway, with crossings running 20–34 a day against 120–147 before the war. For every sector downstream, Hormuz exposure is no longer open or closed but a throughput distribution with fat tails — harder to price than a binary, and probably more durable, because an administrative instrument outlasts a missile exchange. The reopening now needs agreement not merely to stop shooting but on whose flag guarantees the corridor, a harder negotiation than the one that produced the last memorandum. Meanwhile the toll has climbed from flag of convenience hulls to a Gulf state's own tankers, and a second theatre has hardened: the Azov campaign struck about 105 vessels in 9 days, halting Kerch navigation and pushing the disruption from fuel into grain. One clarification that keeps recurring: container rates are running on the Red Sea and peak season, not on Hormuz. The box trade's gate is the Cape diversion, on its own clock into 2027.

Same tariff clocks as autos, opposite sign — and a relief that lasts days.

Apparel sits outside the Section 232 architecture, which inverts the tariff maths against every other sector facing the same July dates: a bare Section 122 sunset on 24 July would cut apparel costs rather than raise them. But the relief is a window measured in days. The Section 301 forced labour findings complete before that sunset, imposing 10–12.5% across 60 economies and stacking on existing tariffs rather than replacing them — and the textile mechanism inside it is keyed to how much US produced cotton an exporting country buys, which makes cotton sourcing the swing variable for the whole country stack. Nike's own planned step from 10% to 15% in August says the exposed operators expect the relief undone. On the ground, the redistribution story needs correcting: displaced volume from China, India and Bangladesh is spreading across Cambodia, Indonesia, Egypt and Turkey rather than concentrating in Vietnam, whose apparel gain is a modest 1.3% and whose producers run 40–45% above Indonesian costs. No single winner banks the lost share. Bangladesh cut its tariff and won on paper, but the order book never moved — over 20,000 garment jobs gone in the first half. Apparel inflation runs 5.1% y/y, above the general index, on 12% lower import volume: the pass through is arriving on a shrinking base.

The reset landed on a specific bet, and it gets scored in winter.

Tourism sits at the bottom of the restoration clock — it depends on aviation, which depends on insurance and regulatory clearance — so it recovers last and breaks first. The corridor hardened again in mid July, with EASA advising avoidance of the Gulf airspace at all levels and a fresh wave of strikes disrupting flights across the region. What makes this more than a diffuse confidence knock is that it lands on a datable commitment. The global chains guided Middle East revenue per room down about 50% for Q2 and still raised full year global guidance — explicitly on the premise that the region recovers sequentially from Q3. The escalation is precisely the event that puts that premise in doubt, so the late July prints test the second half recovery bet rather than just the Q2 floor. Two things keep the read honest. Summer is Dubai's natural soft season, occupancy running 55–65% even in calm years, so the reset is landing in the quarter that hides it best; the real test is the November to March peak, whose forward book is being made now. And the recovery being interrupted was genuine — traffic held near 3M through early July, and 2025 closed at a record 80.7% occupancy. The tell to watch is the booking window, collapsed to 7–14 days from the usual 3 to 4 weeks: rooms still fill, but nobody is committing early.

Every row that eased on the June memorandum has re-armed; every row that did not move on it still has not. That is the whole sorting principle: supply removals divide by mechanism, not by proximity to the conflict. A war premium unwinds when the shooting stops and returns when it resumes — crude, war risk insurance, petrochemical feedstocks and fertiliser all did both inside 3 weeks. A physically damaged floor does neither. Ras Laffan's helium is turbine gated to roughly 2029 whether or not anyone signs anything, and the acid beneath batteries, tyres, fertiliser and defence now rests on 3 separate sovereign decisions rather than one, so no single de-escalation relaxes it.

Two things are new this cycle. The cushion is thinner: the first Hormuz shock was absorbed by the IEA's 400 million barrel release, and that release has since been drawn down, so the same squeeze now meets a market with less to take it. And one contingency has been tested rather than assumed. The war reopening was precisely the condition under which bromine was expected to be severed, and it was not — the price carries a risk premium, not a cut. An armed row that survives its own trigger tells you more than one that has never faced it.

One row is corrected rather than updated: neon's concentration is Chinese, dating from the 2022 shift out of Ukraine, and was never an Iranian input.

Tier 1 is severe and well established. Tier 2 is severe and underpriced. Tier 3 is where the first two arrive downstream.